R-TF-015-003 Clinical Evaluation Report
Table of contents
- Executive summary
- Scope of the clinical evaluation
- Clinical background and SOTA analysis
- Data sources for the state of the art
- Analysis of articles and guidelines for SOTA description
- Description of the medical condition
- WHO ICD-11 Classification of Dermatological Diseases
- Different approaches to dermatological conditions detection
- Possible risks of the use of software and artificial intelligence in dermatological diagnostics
- Minimization and management of possible risks
- Identification of similar devices
- Valid clinical association
- Valid clinical association of the visible skin structures abnormalities
- Valid clinical association of the International Classification of Diseases (ICD) categories
- Conclusions on Valid clinical association of the visible skin structures abnormalities
- Conclusions on Valid clinical association of the International Classification of Diseases (ICD) categories
- Conclusions on Valid Clinical Association of Legit.Health
- Valid clinical association and State of the Art alignment
- Technical performance
- Device equivalence
- Data generated and held by the manufacturer
- Clinical Data
- Clinical performance
- Post-market surveillance
- Compliance with applicable regulatory requirements
- Requirement of safety (GSPR 1)
- Requirement of performance (GSPR 1)
- Requirement of acceptable benefit/risk ratio (GSPR 2, 6)
- Requirement on acceptability of undesirable side-effects (GSPR 8)
- Requirement on devices that incorporate software or for software that are devices in themselves (GSPR 17.2)
- Answer to specific performance and safety questions
- Conclusions
- Date of next clinical evaluation
- Bibliography
- Qualification of the Responsible Evaluators
- Dates and signatures
Executive summary
This Clinical Evaluation Report (CER) has been prepared in accordance with the requirements of Regulation (EU) 2017/745 (Medical Devices Regulation, MDR) and in line with MEDDEV guideline 2. 7/1 revision 4, Clinical evaluation: Guidance for manufacturers and notified bodies under Directives 93/42/EEC and 90/385/EEC, and MDCG 2020-1 Guidance on Clinical Evaluation (MDR) / Performance Evaluation (IVDR) of Medical Device Software as well as MDCG 2020-13 and MDCG 2020-6 as indicated in the associated Clinical Evaluation Plan (CEP).
This CER is a discussion of the benefit/risk profile of using the device to provide support to health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by:
- Providing quantifiable data on the intensity, count and extent of clinical signs such as erythema, desquamation, and induration, among others.
- Providing an interpretative distribution representation of possible International Classification of Diseases (ICD) categories that might be represented in the pixels content of the image.
The device is classified as a class IIb medical device, and the previous version of the device has been commercialized since 2020. The device is manufactured under a Conformity Assessment based on a Quality Management System in accordance with Chapter I of Annex IX of Regulation (EU) 2017/745 Medical Devices.
Compliance with the applicable general safety and performance requirements (GSPRs) is mainly demonstrated based on three key components (valid clinical association, technical performance, and clinical performance), as described in the Clinical Evaluation Plan (CEP). The available evidence for the device includes safety and performance data sourced from non-clinical data sources such as bench-lab testing, which outlines data generated by the manufacturer in terms of pre-clinical and bench testing, along with design verification testing. Additionally, a specific section of this report (Valid clinical association
) elaborates on the valid clinical association through a structured analysis of the relationship between visible skin structure abnormalities and International Classification of Diseases (ICD) categories, including a comprehensive literature review and data appraisal. Furthermore, clinical performance data is detailed in the section Clinical performance
, covering relevant clinical data identification, risk management, and product-specific performance and safety findings derived from both manufacturer data, which includes post-market clinical investigations and literature review, ensuring adherence to the requirements of safety, performance, acceptable benefit/risk ratio, and minimization of undesirable side effects.
We consider that the bibliographic review on similar devices prove the favourable benefit/risk profile of these products intended to give support to health care providers in the assessment of skin structures, when used for the approved intended purposes.
Performance data assessed confirms the adequacy of claims of use foreseen by the manufacturer, and the device is considered suitable and useful for the intended users.
Upon critical review of the published literature and considering that no safety incidents were reported concerning the use of the product, the manufacturer considers that the risk analysis report does not need to be modified.
The previous generation of the device, marketed since 2020 following the acquisition of its Spanish manufacturing license, has undergone continuous evaluation through post-market activities. Since the device's introduction, 21 contracts have been signed with various customers, including both government-run and for-profit care providers. Over 4,500 reports have been generated by more than 500 practitioners, benefitting over 1,000 patients. Notably, no serious incidents or Field Safety Corrective Actions (FSCA) have been reported during the review period. Furthermore, no significant trends or deviations that would necessitate corrective actions were observed in terms of the device's safety or performance. The Post-Market Clinical Follow-Up (PMCF) evaluation report corroborates these findings, noting that user feedback has remained largely positive, with both primary and secondary healthcare professionals expressing high levels of satisfaction.
Clinical investigations, such as the validation study for the early detection of cutaneous melanoma, further substantiate the device's capabilities, with impressive results like an AUC of 0.842 for melanoma detection. Additional studies, including those focused on dermatological conditions and the optimization of clinical flow, have demonstrated positive improvements in diagnostic accuracy and user satisfaction.
Based on the clinical evaluation performed in this report, the risks associated with the intended purpose are minimised and acceptable when weighed against the benefits of using the device as a medical device used to give support to health care providers in the assessment of skin structures, when used for the approved intended purposes.
Acronyms
Acronyms | Definition |
---|---|
CAPA | Corrective and Preventive Actions |
CEP | Clinical Evaluation Plan |
CER | Clinical Evaluation Report |
CET | Clinical Evaluation Team |
EU/EC | European Union/Community |
FDA | Food and Drug Administration |
FMEA | Failure Modes and Effects Analysis |
FSCA | Field Safety Corrective Actions |
GSPR | General Safety and Performance Requirement |
IFU | Instructions For Use |
MA | Metanalysis |
MEDDEV | MEDical DEVices Documents |
MDR | Medical Devices Regulation |
MDSW | Medical Device Software |
PMCF | Post-market Clinical Follow-up |
PMS | Post-market Surveillance |
PSUR | Periodic Safety Update Report |
RCT | Randomized Controlled Trial |
SR | Systematic Review |
STED | Summary Technical Documentation |
USA | United States of America |
Scope of the clinical evaluation
General details
The present clinical evaluation report (CER) is intended to describe the clinical performance and safety of Legit.Health Plus (hereinafter, "the device") as a medical device software (MDSW) used for the assessment of skin structures, enhancing efficiency and accuracy of care delivery. Even though the product is currently certified as a medical device under the Medical Devices Directive 93/42/EEC (MDD), this CER has been performed following the requirements of Regulation EU 2017/745 (Medical Device Regulation, MDR).
The device is a computational software-only medical device intended to support Health Care Practitioners in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing:
- Providing quantifiable data on the intensity, count and extent of clinical signs such as erythema, desquamation, and induration, among others.
- Providing an interpretative distribution representation of possible International Classification of Diseases (ICD) categories that might be represented in the pixels content of the image.
The device is classified as a Class IIb medical device by Rule 11 of Annex VIII of MDR. The previous generation of the device has been commercialized since 2020.
In addition to the requirements as laid in MDR, the CER has been elaborated in accordance with the guidelines and standards listed in section Applicable standards and guidance documents
. Also, the CER follows the procedure GP-015 Clinical Evaluation
of our QMS.
Clinical Evaluation Plan
In accordance with MDR, and as referred to in Article 61 and Annex XIV (Part A), the present clinical evaluation has been performed following the Clinical Evaluation Plan (R-TF-015-001 Clinical Evaluation Plan
).
Objectives
This clinical evaluation is intended to identify, collect, appraise, and analyse all the available clinical data pertaining to the device in order to confirm compliance with the relevant general safety and performance requirements set up in Annex I of MDR when using the device according to the manufacturer's instructions for use. In particular, the objectives of this clinical evaluation are:
- To ascertain whether enough clinical data is available to comply with the applicable general safety and performance requirements (GSPR).
- To ascertain whether the intended performance is adequately supported by sufficient clinical evidence.
- To verify that all the hazards, information on risk mitigation, and clinically relevant information are consistent with the current product's design and verification and adequately reflected in the information supplied by the manufacturer.
- To ascertain whether any residual risk with clinical impact or undesirable side-effects are adequately addressed from a risk management perspective and acceptable when weighted against the clinical benefits brought about by using the product.
- To ascertain whether residual risks and undesirable side-effects are adequately communicated to the user in the information accompanying the product.
- To serve as reliable input data for effective maintenance of the product's risk management.
- To establish that the clinical benefit/risk balance remains positive and acceptable when the product is used as intended by the manufacturer.
Methodology
The present CER is based on the clinical evaluation of the available clinical data related to the device under evaluation as required by MDR. To this end, all relevant clinical data related to the device has been collected, appraised, and analysed following MEDDEV 2.7/1 rev. 4. The requirements for clinical evaluation are outlined in Article 61 of the MDR (including Annex XIV). Three key components should be taken into account:
- Valid clinical association seeks to establish that there are sound scientific principles underpinning the use of the MDSW in question. For this purpose, a systematic literature search has been performed in order to demonstrate the association between between visible skin structure abnormalities and International Classification of Diseases (ICD) categories.
- Technical performance is the demonstration of the MDSW's ability to accurately, reliably and precisely generate the intended output from the input data. This includes the verification and validation activities.
- Clinical performance is the demonstration of a MDSW's ability to yield clinically relevant output in accordance with the intended purpose. The clinical performance shall be mostly supported by a structured, exhaustive and critical review of the clinical scientific literature relevant to the product. In addition, we consider al available data generated and currently held by us.
All relevant clinical data has been collected and appraised in order to establish the safety and performance of the device and to identify any gaps in clinical evidence to support the benefit/risk profile of the device. Details on the followed methodology can be found in the CEP (R-TF-015-001 Clinical Evaluation Plan
).
Applicable standards and guidance documents
The applicable standards and guidance documents to the present CER are listed below:
- MDR 2017/745: Regulation (EU) 2017/745 of the European Parliament and of the Council of 5 April 2017 on medical devices
- MEDDEV 2.7/1 revision 4: European Commission Guidelines on Medical Devices Clinical Evaluation
- IMDRF/AE WG/N43FINAL:2020: IMDRF terminologies for categorized Adverse Event Reporting (AER): terms, terminology structure and codes
- MDCG 2023-3: Questions and Answers on vigilance terms and concepts as outlined in the Regulation (EU) 2017/745 on medical devices
- IMDRF MDCE WG/N57FINAL:2019: Clinical investigation
- MDCG 2024-5 Guidance on content of the Investigator's Brochure for clinical investigations of medical devices
- MDCG 2024-5 Appendix A: Guidance on the Investigator's Brochure content
- MDCG 2024-3 Guidance on content of the Clinical Investigation Plan for clinical investigations of medical devices
- MDCG 2024-3 Appendix A: Guidance on content of the Clinical Investigation Plan for clinical investigations of medical devices
- Clinical Investigation Plan Synopsis Template
- 2023/C 163/06: Commission Guidance on the content and structure of the summary of the clinical investigation report
- MDCG 2020-10/1 Rev.1
- MDCG 2020-10/2 Rev. 1: Guidance on safety reporting in clinical investigations
- Appendix: Clinical investigation summary safety report form
- MDCG 2020-1: Guidance on clinical evaluation (MDR) / Performance evaluation (IVDR) of medical device software
- MDCG 2020-6: Regulation (EU) 2017/745: Clinical evidence needed for medical devices previously CE marked under Directives 93/42/EEC or 90/385/EEC
- MDCG 2022-21: Guidance on Periodic Safety Update Report (PSUR) according to Regulation (EU) 2017/745 (MDR)
- MDCG 2020-7: Guidance on PMCF plan template
- MDCG 2020-8: Guidance on PMCF evaluation report template
- IMDRF MDCE WG/N65FINAL:2021: Post-Market Clinical Follow-Up Studies
- MDCG 2020-13: Clinical evaluation assessment report template
- IMDRF MDCE WG/N56FINAL:2019: Clinical evaluation
- IMDRF MDCE WG/N55 FINAL:2019: Clinical evidence
- ISO 13485:2016, Adm 11: Quality Management Systems - Regulatory Requirements for Medical Devices
- ISO 14971:2019: Medical devices - Application of Risk Management to Medical Devices
Device description
Manufacturer
Manufacturer data | |
---|---|
Legal manufacturer name | AI Labs Group S.L. |
Address | Street Gran Vía 1, BAT Tower, 48001, Bilbao, Bizkaia (Spain) |
SRN | ES-MF-000025345 |
Person responsible for regulatory compliance | Alfonso Medela, María Diez, Giulia Foglia |
office@legit.health | |
Phone | +34 638127476 |
Trademark | Legit.Health |
Device identification
Information | |
---|---|
Device name | Legit.Health Plus (hereinafter, the device) |
Model and type | NA |
Version | 1.0.0.0 |
Basic UDI-DI | 8437025550LegitCADx6X |
Certificate number (if available) | MDR 792790 |
EMDN code(s) | Z12040192 (General medicine diagnosis and monitoring instruments - Medical device software) |
GMDN code | 65975 |
Class | Class IIb |
Classification rule | Rule 11 |
Novel product (True/False) | FALSE |
Novel related clinical procedure (True/False) | FALSE |
SRN | ES-MF-000025345 |
Intended use
The device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing:
- quantification of intensity, count, extent of visible clinical signs
- interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
Quantification of intensity, count and extent of visible clinical signs
The device provides quantifiable data on the intensity, count and extent of clinical signs such as erythema, desquamation, and induration, among others; including, but not limited to:
- erythema,
- desquamation,
- induration,
- crusting,
- xerosis (dryness),
- swelling (oedema),
- oozing,
- excoriation,
- lichenification,
- exudation,
- wound depth,
- wound border,
- undermining,
- hair loss,
- necrotic tissue,
- granulation tissue,
- epithelialization,
- nodule,
- papule
- pustule,
- cyst,
- comedone,
- abscess,
- draining tunnel,
- inflammatory lesion,
- exposed wound, bone and/or adjacent tissues,
- slough or biofilm,
- maceration,
- external material over the lesion,
- hypopigmentation or depigmentation,
- hyperpigmentation,
- scar,
- ictericia
Image-based recognition of visible ICD categories
The device is intended to provide an interpretative distribution representation of possible International Classification of Diseases (ICD) categories that might be represented in the pixels content of the image.
Device description
The device is a computational software-only medical device leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures. Its principal function is to provide a wide range of clinical data from the analyzed images to assist healthcare practitioners in their clinical evaluations and allow healthcare provider organisations to gather data and improve their workflows.
The generated data is intended to aid healthcare practitioners and organizations in their clinical decision-making process, thus enhancing the efficiency and accuracy of care delivery.
The device should never be used to confirm a clinical diagnosis. On the contrary, its result is one element of the overall clinical assessment. Indeed, the device is designed to be used when a healthcare practitioner chooses to obtain additional information to consider a decision.
Intended medical indication
The device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
Intended patient population
The device is intended for use on images of skin from patients presenting visible skin structure abnormalities, across all age groups, skin types, and demographics.
Intended user
The medical device is intended for use by healthcare providers to aid in the assessment of skin structures.
User qualification and competencies
In this section we specificy the specific qualifications and competencies needed for users of the device, to properly use the device, provided that they already belong to their professional category. In other words, when describing the qualifications of HCPs, it is assumed that healthcare professionals (HCPs) already have the qualifications and competencies native to their profession.
Healthcare professionals
No official qualifications are needes, but it is advisable if HCPs have some competencies:
- Knowledge on how to take images with smartphones.
IT professionals
IT professionals are responsible for the integration of the medical device into the healthcare organisation's system.
No specific official qualifications are needed, but it is advisable that IT professionals using the device have the following competencies:
- Basic knowledge of FHIR
- Understanding of the output of the device.
Use environment
The device is intended to be used in the setting of healthcare organisations and their IT departments, which commonly are situated inside hospitals or other clinical facilities.
The device is intended to be integrated into the healthcare organisation's system by IT professionals.
Operating principle
The device is computational medical tool leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures.
Body structures
The device is intended to use on the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
In fact, the device is intended to use on visible skin structures. As such, it can only quantify clinical signs that are visible, and distribute the probabilities across ICD categories that are visible.
Contraindications and precautions required by the manufacturer
Contraindications
We advise not to use the device if:
- Skin structures located at a distance greater than 1 cm from the eye, beyond the optimal range for examination.
- Skin areas that are obscured from view, situated within skin folds or concealed in other manners, making them inaccessible for camera examination.
- Regions of the skin showcasing scars or fibrosis, indicative of past injuries or trauma.
- Skin structures exhibiting extensive damage, characterized by severe ulcerations or active bleeding.
- Skin structures contaminated with foreign substances, including but not limited to tattoos and creams.
- Skin structures situated at anatomically special sites, such as underneath the nails, requiring special attention.
- Portions of skin that are densely covered with hair, potentially obstructing the view and hindering examination.
Precautions
After analysing the risks associated to the use of the device, we have identified some residual risks.
The following table summarizes the residual risks and the recomended course of action for each of them:
# | Situation | Recommended course of action |
---|---|---|
5 | Incorrect clinical information: the care provider receives into their system data that is erroneous | The device must always be used under the supervision HCP, who should confirm or validate the output of the device considering the medical history of the patient, and other possible sympthoms they could be suffering, especially those that are not visible or have not been supplied to the device |
6 | Incorrect diagnosis or follow up: the medical device outputs a wrong result to the HCP | The device must always be used under the supervision HCP, who should confirm or validate the output of the device considering the medical history of the patient, and other possible sympthoms they could be suffering, especially those that are not visible or have not been supplied to the device. Also, we encourage you to review the metadata returned by the device about the output, such as explainability media and other metrics. |
9 | Image artefacts/resolution: the medical device receives an input that does not have sufficient quality in a way that affects its performance | The Instructions for Use contain extensive indication on how to take pictures in a section called 'How to take pictures'. We also offer training to the users to improve the imaging process so that it is optimal for the device's operation; feel free to request such training to your closest sales representative. Also, we encourage you to pay attention to the information regarding image quality that the device outputs alognside the clinical information. |
11 | Data transmission failure from care provider's system: the care provider's system cannot connect to the device to send data | The Instructions for Use contain extensive indication on how to integrate the device into the care provider's system in a section called 'Installation manual'. |
12 | Data input failure: the medical device cannot receive data from care providers | The Instructions for Use contain extensive indication on how to integrate the device into the care provider's system in a section called 'Installation manual'. |
13 | Data accessibility failure: the care provider cannot receive data from the medical device | The Instructions for Use contain extensive indication on how to integrate the device into the care provider's system in a section called 'Installation manual'. |
14 | Data transmission failure: the medical device cannot send data to care providers | The Instructions for Use contain extensive indication on how to integrate the device into the care provider's system in a section called 'Installation manual'. |
30 | Inadequate lighting conditions during image capture: The medical device receives an input that does not have sufficient quality | This is similar to risk ID 9, but with a very simple solution: use the flash. If you can't use the flash and still the image is dark, move to a different environment with better lightning. Also, we encourage you to pay attention to the information regarding image quality that the device outputs alognside the clinical information. |
Warnings
In case of observing an incorrect operation of the device, notify us as soon as possible. You can use the email support@legit.health. We, as manufacturers, will proceed accordingly. Any serious incident should be reported to us, as well as to the national competent authority of the country.
Undesirable effects
Any undesirable side-effect should constitute an acceptable risk when weighed against the performances intended.
It is not known or foreseen any undesirable side-effects specifically related to the use of the software.
Instructions for Use
The IFU of the device are developed according to the applicable requirement of MDR 2017/745, Annex I. As indicated in the IFU document, the use methodology is as follows:
The device is architected to seamlessly integrate with other software platforms. Primarily designed as an Application Programming Interface (API), it allows healthcare organizations to establish a real-time connection between their native systems, such as Electronic Medical Records (EMR) systems, and the device. This ensures that images can be sent from the EMR and clinical data from the device can be received and stored back into the EMR in real time.
Our instructions for use can be found at Legit.Health Plus_IFU
and are made available to the users in our webpage; they contain detailed and helpful information to help our client integrate the device into their systems.
The IFU include relevant information such as intended use, warnings or contra-indications, which have been included in the Device identification
section above.
Components
The device is a computational software-only medical device leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures.
Variants
No variants.
Accessories of the product
- Primary accessories are the components that interact directly with the device. These can be known by the manufacturer. They are also required to interact with the device. The device is used through an API (Application Programming Interface). This means that the interface is coded, and used programmatically, without a user interface. In other words: the device is used server-to-server, by computer programs. Thus, no accessory is used directly in interaction with the device.
- Secondary accessories are the components that may interact indirectly with the device. These are developed and maintained independently by the user, and the manufacturer has no visibility as to their identity or operating principles. They are also optional and not required to interact with the device.
The device may also be used indirectly through applications, such as the care provider's Electronic Health Records (EHR). The EHR is the software system that stores patients' data: medical and family history, laboratory and other test results, prescribed medications history, and more. This is developed and maintained independently of us, and may be used to indirectly interact with the device.
The patients and healthcare providers may use image capture devices to take photos of skin structures. In this regard, the minimum requirement is a 12 MP camera.
Device materials in contact with patient or user
The device is intended to analyse images of the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis). Due to the nature of the device (stand-alone software), it doesn't come into contact with tissue or bodily fluids.
Technical specifications
API REST
Our device is built as an API that follows the REST protocol.
This protocol totally separates the user interface from the server and the data storage. Thanks to this, REST API always adapts to the type of syntax or platforms that the user may use, which gives considerable freedom and autonomy to the user. With a REST API, the user can use either PHP, Java, Python or Node.js servers. The only thing is that it is indispensable that the responses to the requests should always take place in the language used for the information exchange: JSON.
OpenAPI Specification
Our medical device includes an OpenAPI Specification.
OpenAPI Specification (formerly known as Swagger Specification) is an API description format for REST APIs. An OpenAPI file allows you to describe a entire API, including:
- Available endpoints and operations on each endpoint (GET, POST)
- Operation parameters Input and output for each operation
- Authentication methods
- Contact information, license, terms of use and other information.
This means that our API itself has embedded specifications that help the user understand the type of values that are transmitted by the API.
HL7 FHIR
FHIR is a standard for health care data exchange, published by HL7®. FHIR is suitable for use in a wide variety of contexts: mobile phone apps, cloud communications, EHR-based data sharing, server communication in large institutional healthcare providers, and much more.
FHIR solves many challenges of data interoperability by defining a simple framework for sharing data between systems.
The relevant performance attributes of the devices are described in the following table.
Metric | Value |
---|---|
Weight | 33 kilobytes |
Average response time | 1400 miliseconds |
Maximum requests per second | No limit |
Service availability time slot | The service is available at all times |
Service availability rate during its working slot (in % per month) | 100% |
Maximum application recovery time in the event of a failure (RTO/AIMD) | 6 hours |
Maximum data loss in the event of a fault (none, current transaction, day, week, etc.) (RPO/PDMA) | None |
Maximum response time to a transaction | 10 seconds |
Backup device (software, hardware) | Software (AWS S3) |
Backup frequency | 12 hours |
Backup modality | Incremental |
Recomended dimensions of images sent | 10,000px2 |
How the device achieves its intended purpose
Principle of operation
The device is a computational medical tool leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures.
Mode of action
One core feature of the device is a deep learning-based image recognition technology for the recognition of ICD categories. In other words: when the device is fed an image or a set of images, it outputs an interpretative distribution representation of possible International Classification of Diseases (ICD) categories that might be represented in the pixels content of the image.
The device makes its prediction entirely based on the visual content of the images, with no additional parameters.
The device has been developed following an architecture called Vision Transformer (ViT). This architecture is inspired in the Transformer architecture, which is extensively used in other areas such as NLP and has brought significant advancements in terms of performance.
Another core feature of the device is to provide a quantifiable data on the intensity, count and extent of clinical signs such as erythema, desquamation, and induration, among others.
To achieve that, the device uses a range of deep learning technologies, combined and developed for that specific use. Here's a list of the technologies used:
- Object detection: used to count clinical signs such as hives, papules or nodules.
- Semantic segmentation: used to determine the extent of clinical signs such as hair loss or erythema.
- Image recognition: used to quantity the intensity of visual clinical signs like erythema, excoriation, dryness, lichenification, oozing, and edema.
Use environment
The device is intended to be used in the setting of healthcare organisations and their IT departments, which commonly are situated inside hospitals or other clinical facilities.
Clinical benefits
The device provides clinical benefits by enhancing the precision and efficiency of dermatological assessments through advanced image analysis of visible skin structures. By quantifying the intensity, count, and extent of clinical signs, it offers detailed and consistent data, which aids healthcare providers in evaluating a wide range of skin conditions across the epidermis, appendages, associated mucous membranes, dermis, cutaneous vasculature, and subcutis. Additionally, it interprets and maps possible ICD classifications, streamlining clinical workflows and supporting standardized assessments for more accurate, data-driven patient care.
Claims or intended claims
- Claim 1: Access to comprehensive dermatological data for more informed and precise clinical decision-making.
- Claim 2: Enhanced accuracy in dermatological assessments.
- Claim 3: Faster quantification of clinical signs.
- Claim 4: Precise, objective quantification of the intensity, quantity, and extent of clinical signs.
- Claim 5: More efficient monitoring of skin conditions, with consistent and reliable tracking of patient progress.
Data collection, model training and validation
In order to develop this image recognition feature, we created an image dataset by collecting images from diverse sources, including the most famous skin image datasets, which included the corresponding confirmed ICD category for every image. This results in a dataset of near 1000 different classes and a diverse representation of age, sex and skin tone. To ensure the model has enough images per category, only classes with a sufficient number of images are considered for training the model.
One crucial step of the development is splitting the dataset into three groups:
- Training sets
- Validation sets
- Test sets
When an incoming image dataset includes any sort of metadata that makes it possible to group images by subject, the data is split at subject level. This strategy improves the reliability of the validation and test metrics, and it is a best practice in the field.
Thanks to such a large collection of datasets, it is also possible to perform reliable validations by reserving some of these datasets entirely for testing, which helps explore and analyze the performance of the model in completely uncontrolled scenarios.
Model calibration
Image recognition models are known for becoming overly confident after training, which, applied to this scenario, can deter the overall ICD classification performance. If the device predicts the wrong probabilities and sets an extremely high probability for an incorrect class, the answer may lead the user to believe the model is confident about the answer.
To overcome this problem, a final step has been added to the training process, which consists of slightly fitting the model to the validation set, applying temperature scaling. By applying this additional postprocessing, the device is enforced to generate 'softer' or less extreme probability distributions.
The main benefit of model calibration is that it increases the interpretability of the results, as models not only need to be accurate (i.e. the correct class should be within the top 3 or 5 classes with the highest probability in the distribution) but also indicate how likely it is that the output is not correct. When the model is fed with an image and the output is a very soft distribution, it means that the model is not confident enough about any particular ICD category.
A more detailed explanation can be found in the publication "On calibration of modern neural networks" (Guo et al., 2017).
Status of commercialization
This product has not been commercialized yet. It is undergoing initial CE mark.
Previous version of the device
The predecessor of the current device is named Legit.Health. This earlier version was designed to provide a standalone interface as well as an API, allowing users to engage with it independently of their existing Electronic Health Records. However, this design direction was later recognized as suboptimal, as most organizations expressed a preference for integrating the device's functionalities directly into their existing systems.
In other words: the new generation differs from the previous device in that it's meant to integrate into organisation's softwares. It is used server-to-server, by computer programs. This means that the new device is simpler and contains less elements. This design allows us to focus on other issues, such as interoperability and documentation. And it allows us to invest our development efforts into the scalable architecture of the device, the structure of the input and the output, and helping customers during the integration process.
Legit.Health has been commercialized since 2020 (after obtaining the manufacturing license in Spain) and was certified under the Medical Devices Directive (MDD). Since that time, we've partnered with 21 diverse customers. The customers span from goverment-run care providers to for-profit care providers. Over this period, more than 4,500 diagnostic reports have been crafted by more than 500 professionals. They've utilized our product to help over a thousand patients.
Clinical background and SOTA analysis
Data sources for the state of the art
In the context of the European Union's Medical Devices Regulation (MDR) 2017/745, state of the art refers to the current level of technical development and accepted clinical practice in products, processes and patient management. Although it lacks an explicit definition, it is understood as the consolidated state of knowledge in science and technology at a specific point in time. It does not necessarily imply the most advanced, expensive or frequently used solution, but what is currently accepted as good practice. The identification and understanding of this state is crucial for risk assessment and plays a key role in the writing of clinical evaluation reports, ensuring alignment with the intended use of medical devices and effective management of associated risks.
In relation to the current knowledge/ state of the art in the relevant medical field, the following aspects and information have been verified:
- Applicable standards and medical guidelines.
- Information related to the medical condition managed with the device.
- Other similar devices marketed and medical alternatives available for the target population.
Strategy for literature search
On 23rd October 2024 a systematic literature search was performed in the databases of PubMed with the objective to find relevant information on safety and efficacy of the device under evaluation.
The algorithm used in PubMed was:
("skin cancer" OR "epidermis" OR "chronic skin conditions" OR "skin conditions" OR "inflammatory skin diseases" OR "malignant skin lesions" OR "melanoma" OR "acne" OR "psoriasis" OR "dermatofibroma" OR "dermatosis") AND ("software" OR "digital imag\*" OR "smartphone" OR "web application") AND ("SkinVision" OR "artificial intelligence" OR "machine learning" OR "deep learning" OR "computer vision" OR "deep neural networks" OR "metaoptima" OR "clinical exam" OR "visual inspection" OR "manual assessment") AND ("estimation" OR "classification" OR "followup" OR "diagnos\*" OR "quantif\*")
The literature search was performed following the PICO method described in the Clinical Evaluation Plan. As a result of the search, 118 articles were retrieved and screened by the title and abstract's content.
The 118 articles are listed in Attachment 02: Literature Search Records in the SOTA folder.
62 articles were discarded as they were not related to the device under evaluation or the medical condition. The remaining 56 articles were appraised according to the criteria for appraisal defined in the Clinical Evaluation Plan.
Finally, 12 articles from PubMed were selected for their evaluation as part of the state of the art. A summary of the whole process in described in the following flowchart (Flowchart of the systematic SOTA literature search
). The analysis of the 12 articles is provided in the Analysis of articles and guidelines for SOTA description
section.
Applicable standards and guidance documents
The clinical evaluation of the device will be performed according to the relevant legal framework and following the applicable and established standards described in the Clinical Evaluation Plan.
A literature search of guidelines has been performed in Google and Pubmed searching the following terms: ICD-11 disease of skin guideline in order to find medical guidelines related with ICD-11 Classification of Dermatological Diseases.
The following guidance document has been considered for the establishment of SOTA:
- The WHO ICD-11 Classification of Dermatological Diseases: a new comprehensive online skin disease taxonomy designed by and for dermatologists. British Journal of Dermatology. 2022
The analysis of the guideline is provided in Guidance documents
section.
Similar devices
A comparison of the main technical and clinical characteristics with other products that similar and available on the market has been performed in order to ensure its alignment with the current standards.
The comparison is provided in Identification of similar devices
section.
Analysis of articles and guidelines for SOTA description
Articles from systematic literature review
Article ID 12
- Title / Year: Evaluation of an artificial intelligence-based decision support for detection of cutaneous melanoma in primary care - a prospective, real-life, clinical trial (2024)
- Author: Papachristou, P.
- DOI / PMID: 10.1093/bjd/ljae021
- Brief summary of the article: the study investigates the performance of an AI-based clinical decision support tool, Dermalyser®, used by primary care physicians (PCPs) for melanoma detection through smartphone app technology. Early detection and excision of melanoma are critical for improving prognosis and survival, yet differentiating between melanoma and benign lesions can be challenging, even for experienced dermatologists. The app leverages machine learning algorithms trained on extensive datasets of dermoscopic images, achieving an area under the receiver operating characteristic (ROC) curve (AUROC) of 0.94 in pre-study simulations. In a prospective clinical trial across 36 primary care centers in southern Sweden, 138 PCPs assessed 253 skin lesions with varying degrees of melanoma suspicion. The app provided binary outcomes indicating the likelihood of melanoma, with a predefined cutoff ensuring a balance between sensitivity (95.2%) and specificity (60.3%). The app demonstrated strong diagnostic accuracy, achieving an AUROC of 0.960 for all melanoma cases and 0.988 for invasive melanomas, alongside a notable negative predictive value (NPV) of 99.5%. This suggests that the AI tool can significantly assist PCPs in making more informed decisions regarding suspicious lesions, potentially reducing unnecessary referrals and excisions while enhancing overall diagnostic confidence. The study underscores the need for further large-scale research in primary care settings to fully validate AI's effectiveness in real-world scenarios, particularly as many patients first present their skin concerns to PCPs.
- IFU claims interpretation:
- Intended Purpose: the device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: the device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
- Indications: the device is indicated for use on images of visible skin structure abnormalities for supporting the assessment of all skin diseases listed and described in the ICD-11 (code 14).
- Operating Principle: The device is a computational medical tool leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures.
Article ID 25
- Title / Year: A novel Skin lesion prediction and classification technique: ViT-GradCAM (2024)
- Author: Shafiq, M.
- DOI / PMID: 10.1111/srt.70040
- Brief summary of the article: The article discusses the development and implementation of the ViT-GradCAM model, a novel deep learning approach designed for assessing skin conditions through image analysis. Skin lesions, categorized into primary and secondary types, can indicate serious conditions such as melanoma, a type of skin cancer whose incidence is rising globally. Effective classification of these lesions is crucial for early diagnosis and treatment.
The ViT-GradCAM model utilizes a vision transformer (ViT) to analyze skin images by breaking them into patches for feature extraction. To enhance visualization and mitigate information loss during classification, Gradient-Weighted Class Activation Mapping (GradCAM) is integrated. This approach addresses the challenges of data scarcity in skin condition datasets by employing techniques like data augmentation and synthetic data generation to improve model training.
Key features of the ViT-GradCAM model include:
- Data Preprocessing and Augmentation: Enriching spatial feature extraction to enhance model performance.
- Class Imbalance Mitigation: Generating artificial data to improve learning outcomes
- Gradient-Weighted Backpropagation: Preserving spatial information during the final classification layer.
- Real-Time Diagnosis Tool: Providing a user-friendly interface for medical professionals to classify and diagnose skin lesions quickly. The model achieved an impressive classification accuracy of approximately 96.6% across seven different skin lesion categories, demonstrating superior performance compared to conventional algorithms. The study highlights the importance of leveraging deep learning techniques, particularly those that enhance data visualization and model robustness, to improve the accuracy of skin lesion assessments.
- IFU claims interpretation:
- Intended purpose: the device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: the device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
- Contraindications: We advise not to use the device in: Skin structures located at a distance greater than 1 cm from the eye, beyond the optimal range for examination. Skin areas that are obscured from view, situated within skin folds or concealed in other manners, making them inaccessible for camera examination. Regions of the skin showcasing scars or fibrosis, indicative of past injuries or trauma. Skin structures exhibiting extensive damage, characterized by severe ulcerations or active bleeding. Skin structures contaminated with foreign substances, including but not limited to tattoos and creams. Skin structures situated at anatomically special sites, such as underneath the nails, requiring special attention. Portions of skin that are densely covered with hair, potentially obstructing the view and hindering examination.
- Part of the body: the device is intended to analyse images of the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
Article ID 46
- Title / Year: Design and validation of a new machine-learning-based diagnostic tool for the differentiation of dermatoscopic skin cancer images (2023)
- Author: Tajerian, A.
- DOI / PMID: 10.1371/journal.pone.0284437
- Brief summary of the article: The integration of artificial intelligence (AI) and machine learning (ML) into dermatology is revolutionizing the diagnosis of skin cancer. Skin cancer remains a critical health issue globally, with rising incidence rates, particularly among fair-skinned populations vulnerable to UV radiation damage. Traditional diagnostic methods often require histopathological examinations, which can delay timely treatment. However, AI-driven technologies, particularly those utilizing deep learning algorithms, are paving the way for faster and more accurate assessments. One significant dataset used for developing AI diagnostic tools is the HAM10000 dataset, containing 10,015 images of pigmented skin lesions categorized into seven subgroups, including melanoma, basal cell carcinoma (BCC), and benign keratosis. The dataset is crucial for training machine-learning algorithms to classify skin lesions effectively. A study demonstrated that state-of-the-art machine-learning classifiers could outperform human dermatologists, achieving a mean improvement of 2.01 correct diagnoses per 30-image batch. Researchers have trained deep learning models, such as the EfficientNetB1, on the HAM10000 dataset, achieving an impressive accuracy of 84.3% in classifying dermatoscopic images of skin lesions. This model utilized transfer learning and fine-tuning techniques to enhance its performance, making it a valuable tool for dermatologists in clinical practice. The algorithm identifies patterns in skin lesions that may be invisible to the human eye, facilitating informed decisions and tailored treatments. In a notable study by Esteva et al. at Stanford University, a deep learning model was developed using over 129,000 images of skin lesions, achieving an area under the curve (AUC) of 0.91. This suggests that AI can significantly enhance diagnostic accuracy, potentially providing affordable and accessible diagnostic solutions through smartphone applications. The model's development involved extensive training on a robust infrastructure, including high-performance GPUs, and demonstrated strong performance metrics, such as high precision and recall, particularly in detecting melanocytic nevi. Future work aims to integrate this technology into user-friendly applications, further improving dermatological diagnostics and patient care.
- IFU claims interpretation:
- Intended purpose: the device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: the device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
- Indications: the device is indicated for use on images of visible skin structure abnormalities for supporting the assessment of all skin diseases listed and described in the ICD-11 (code 14).
- Operating principle: The device is a computational medical tool leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures.
- Part of the body: the device is intended to analyse images of the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
Article ID 49
- Title / Year: Multiclass skin lesion localization and classification using deep learning based features fusion and selection framework for smart healthcare (2023)
- Author: Maqsood, S.
- DOI / PMID: 10.1016/j.neunet.2023.01.022
- Brief summary of the article: The healthcare industry is increasingly utilizing advanced technologies, particularly software and artificial intelligence (AI), to assess skin conditions through image analysis. Due to a shortage of medical professionals and resources, automated detection methods are being developed to improve the accuracy and efficiency of skin cancer diagnoses. Melanoma, a particularly aggressive skin cancer, has seen a dramatic rise in incidence, making early detection crucial for improving survival rates. Traditional methods of skin examination, including visual inspections and dermoscopy, can be time-consuming and subject to human error. To address these challenges, researchers are focusing on computer-aided diagnostics (CAD) that leverage deep learning and convolutional neural networks (CNN) to automatically identify skin lesions in dermoscopic images. These systems streamline the diagnostic process by integrating image acquisition, feature extraction, segmentation, and classification into a cohesive workflow. CAD approaches aim to enhance diagnostic accuracy by extracting deep features from images, enabling more reliable differentiation between benign and malignant lesions. Advanced segmentation techniques help isolate lesions for analysis, though challenges remain, such as low-contrast images and variability in lesion characteristics. Recent developments include custom CNN architectures designed specifically for skin lesion segmentation, as well as methods that improve image quality through contrast enhancement. These innovations represent significant progress in the early detection of skin cancer, ultimately aiming to support healthcare professionals in making informed clinical decisions.
- IFU claims interpretation:
- Intended purpose: the device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: the device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
- Operating principle: The device is a computational medical tool leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures.
- Part of the body: the device is intended to analyse images of the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
Article ID 56
- Title / Year: Lesion identification and malignancy prediction from clinical dermatological images (2022)
- Author: Xia, M.
- DOI / PMID: 10.1038/s41598-022-20168-w
- Brief summary of the article: The study discusses a software and artificial intelligence (AI) approach designed to assess skin conditions through image evaluation. Prior to the COVID-19 pandemic, dermatology care faced significant access challenges, characterized by long waiting times for appointments and an increasing incidence of skin cancers. The pandemic exacerbated these issues, as dermatology consultations dropped by 80-90%, delaying care for urgent dermatologic concerns. The aging population, particularly the growing Medicare demographic, is at heightened risk for skin cancer, necessitating innovative solutions to enhance access to dermatology services. The proposed AI-based framework aims to address these access issues by utilizing a two-stage approach that detects skin lesions in images captured using consumer-grade smartphones. This approach includes binary classification to categorize lesions as malignant or benign, targeting common skin cancers like melanoma, basal cell carcinoma, and squamous cell carcinoma, as well as other benign tumors. Ground truth malignancy is determined via biopsy, which strengthens the model's clinical relevance and applicability to primary care and dermatology workflows. Unlike existing methods, this framework is capable of analyzing both wide-field clinical images and dermoscopy images obtained from smartphones, providing a more accessible tool for healthcare providers. The research evaluates various design choices for the AI model, analyzing metrics such as Area Under the Curve (AUC) and Average Precision (AP) to assess performance. Results indicate that the model performs slightly better on dermoscopy images, attributed to their higher quality and resolution. The study also explores the model's efficacy in automatic lesion detection, demonstrating strong performance metrics across different scenarios, particularly for the one-class model, which excels in identifying true lesions. In addition to image-based predictions, the integration of clinical covariates was examined, showing minimal improvements when combined with image data. This underscores the predictive power of images in diagnosing malignancies.
- IFU claims interpretation:
- Intended purpose: the device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: the device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
- Operating principle: The device is a computational medical tool leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures.
- Part of the body: the device is intended to analyse images of the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
Article ID 57
- Title / Year: AcneGrader: An ensemble pruning of the deep learning base models to grade acne (2022)
- Author: Liu, S.
- DOI / PMID: 10.1111/srt.13166
- Brief summary of the article: The study discusses the use of advanced software and AI, particularly graph neural networks (GNNs), for assessing acne severity through image analysis. Acne vulgaris, prevalent among adolescents and adults, significantly impacts individuals' quality of life and self-esteem. Traditionally, dermatologists use the Hayashi criterion to grade acne severity into four levels based on lesion counts. The researchers propose a novel classification framework, called AcneGrader, which employs an ensemble of deep learning models to improve accuracy in grading acne. The framework includes three main components:
- Base Model Training and Pruning: Multiple base models are trained, and an ensemble pruning method is employed to eliminate redundant models based on their Kappa statistic, which reflects model diversity. This step ensures that only the most diverse and effective models are retained.
- Feature Representation and Selection: The predictions from the selected base models are treated as new features for the ensemble framework. Feature selection algorithms further refine this feature set by removing less impactful features.
- Classifier for Ensemble: A classifier combines the results of the selected base models, allowing for the assignment of different weights to each model based on their contributions. The study highlights the importance of ensemble algorithms, such as random forests, and the challenges of computational complexity. It shows that an ensemble pruning strategy can enhance model efficiency without sacrificing accuracy. The experimental results indicate that AcneGrader outperforms existing methods in classifying acne severity, achieving a top accuracy of 93.18%. In summary, the AcneGrader framework utilizes deep learning and ensemble pruning techniques to improve the assessment of acne through image analysis, addressing the complexities of traditional grading methods while significantly enhancing predictive performance.
- IFU claims interpretation:
- Intended purpose: the device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: the device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
- Operating principle: The device is a computational medical tool leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures.
- Part of the body: the device is intended to analyse images of the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
Article ID 68
- Title / Year: Validation of a Market-Approved Artificial Intelligence Mobile Health App for Skin Cancer Screening: A Prospective Multicenter Diagnostic Accuracy Study (2022)
- Author: Sangers, T.
- DOI / PMID: 10.1159/000520474
- Brief summary of the article: The study examines the effectiveness of an mHealth app, SkinVision, which employs a deep-learning convolutional neural network (CNN) to assess skin lesions for premalignancy and malignancy. Conducted in Dutch dermatology clinics, the prospective cross-sectional study involved 785 skin lesions from 372 patients, using smartphones with high-resolution cameras to capture images. The app processes these images to classify lesions as low or high risk for skin cancer within 30 seconds. Deep learning algorithms have shown promise in achieving diagnostic accuracy comparable to dermatologists, highlighting the potential for improved skin cancer detection, particularly in closed healthcare systems where general practitioners may have limited diagnostic capabilities. Despite the advancement of mobile applications, a systematic review indicated a lack of robust evidence supporting the accuracy of such tools, emphasizing the need for prospective validation studies. The results showed that the app successfully identified 239 of 275 premalignant and malignant lesions, yielding an overall sensitivity of 86.9% and specificity of 70.4%. Sensitivity varied by device type, with higher accuracy reported for iOS devices compared to Android. Furthermore, lesions in skin folds demonstrated higher sensitivity but lower specificity than those on smooth skin, suggesting the environment can affect algorithm performance. In conclusion, while the mHealth app demonstrated promising diagnostic capabilities, further validation and research are necessary to establish its clinical utility in routine dermatological practice.
- IFU claims interpretation:
- Intended purpose: the device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: the device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
- Operating principle: The device is a computational medical tool leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures.
- Part of the body: the device is intended to analyse images of the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
Article ID 72
- Title / Year: Implementation of artificial intelligence algorithms for melanoma screening in a primary care setting (2021)
- Author: Giavina-Bianchi, M.
- DOI / PMID: 10.1371/journal.pone.0257006
- Brief summary of the article: The implementation of artificial intelligence (AI) algorithms in melanoma screening aims to enhance early detection of skin cancers in primary care settings, particularly in Brazil, where dermatologists are scarce. The study developed a computer-aided diagnosis (CAD) system that includes two models: one for dermoscopy images and another for clinical images taken with smartphones. The dermoscopy model was trained on a dataset comprising 26,342 images, supplemented with 4,000 artificially generated melanoma images to address class imbalance. This model employs a convolutional neural network (CNN) architecture, specifically using an ensemble of EfficientNetB6 models, to classify lesions based on their likelihood of being melanoma. For clinical images, which are generally of lower quality, the system uses a similarity network to filter out images that don't meet quality standards. This model identifies “ideal” images and categorizes others based on artifacts present. A VGG16 network is utilized to extract features, which are then classified with a K-Nearest Neighbor algorithm. The clinical model is designed to output a binary classification indicating whether a lesion is suspicious for malignancy. To improve user understanding, the CAD system integrates a HeatMap generated through Grad-CAM, highlighting which areas of the image influenced the AI's decision-making. This visual explanation assists primary care physicians (PCPs) in interpreting AI outputs and integrating them into clinical workflows, thereby enhancing patient management and early detection of skin cancers. The overarching goal is to provide an efficient tool that supports PCPs in diagnosing melanoma and potentially expanding to other keratinocytic cancers in the future.
- IFU claims interpretation:
- Intended purpose: the device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: the device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
- Operating principle: The device is a computational medical tool leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures.
- Part of the body: the device is intended to analyse images of the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
Article ID 77
- Title / Year: Automated psoriasis lesion segmentation from unconstrained environment using residual U-Net with transfer learning (2021)
- Author: Raj, R.
- DOI / PMID: 10.1016/j.cmpb.2021.106123
- Brief summary of the article: The proposed research focuses on developing an automated method for assessing psoriasis lesions using deep learning techniques applied to image analysis. Psoriasis, a chronic autoimmune skin condition affecting approximately 125 million people globally, presents significant challenges for accurate diagnosis and treatment due to its rapid skin cell turnover and varied clinical manifestations. A crucial aspect of psoriasis management is the assessment of lesion severity, commonly evaluated through the Psoriasis Area Severity Index (PASI). Traditional clinical methods require significant expertise and time, prompting a shift towards more automated solutions using advanced technologies like artificial intelligence (AI) and deep learning. Recent advancements in deep learning have shown promise in medical image analysis, particularly for skin conditions. The study utilizes the U-Net architecture, known for its efficacy in image segmentation tasks, to segment psoriasis lesions from complex background images. This approach incorporates transfer learning techniques, leveraging a pre-trained residual network within the U-Net model to enhance segmentation accuracy, especially given the limited availability of labeled training data. Key contributions of the research include:
- Automated Segmentation: The study introduces a fully automated deep learning framework capable of segmenting multiple psoriasis lesions from raw digital images, effectively managing variations in background complexity.
- Data Acquisition: A diverse dataset comprising 500 images from 100 psoriasis patients was collected in an unconstrained environment, reflecting real-world conditions without artificial enhancements.
- Transfer Learning and Training Strategies: The effectiveness of transfer learning is explored, demonstrating improved performance with various training strategies and input image sizes, which is crucial given the limited data available for psoriasis lesion segmentation.
- Performance Evaluation: The proposed model is rigorously compared against established deep learning frameworks (e.g., FCN, SegNet, U-Net, PsLSNET) and shows statistically significant improvements in segmentation metrics, including Dice Index (DI) and Jaccard Index (JI).
- IFU claims interpretation:
- Intended purpose: the device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: the device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
- Part of the body: the device is intended to analyse images of the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
Article ID 82
- Title / Year: Performance of a deep neural network in teledermatology: a single-centre prospective diagnostic study (2020)
- Author: Muñoz-López, C.
- DOI / PMID: 10.1111/jdv.16979
- Brief summary of the article: The use of artificial intelligence (AI) in diagnosing skin diseases has shown significant potential to enhance health outcomes. Recent advancements in AI algorithms, particularly those utilizing convolutional neural networks (CNNs), demonstrate the capability to classify clinical images with accuracy comparable to or superior to that of dermatologists in controlled settings. However, these algorithms have yet to be widely implemented in clinical dermatology and have limited real-world efficacy, with only one study assessing their performance in practical settings. During the COVID-19 pandemic, a dermatology service adopted a live, video-based telemedicine approach, emphasizing the need for accurate AI tools to support diagnostics amid telemedicine limitations. A notable study involved a publicly accessible CNN algorithm trained on over 220,000 images of 174 dermatological conditions, which was validated against a diverse dataset of images. In a prospective study conducted via telemedicine, patients submitted images of their skin conditions, which dermatologists assessed during video consultations. The AI algorithm provided three probable diagnoses ranked by likelihood, and its utility was evaluated based on its impact on diagnostic confidence and clinical decision-making. Results indicated that while the AI algorithm's top-1 accuracy was lower than that of dermatologists, its balanced top-1 accuracy—accounting for varying condition prevalence—was comparable to that of dermatologists and superior to general practitioners. The study's unique aspects included evaluating patient-taken images in real-life teledermatology settings, emphasizing the algorithm's potential role as a decision-support tool. However, challenges such as the algorithm's performance being affected by image quality and skin type variability were noted, highlighting the importance of robust image quality standards and diverse training datasets for AI applications in dermatology. Overall, AI algorithms present a promising avenue for enhancing the diagnostic capabilities of general practitioners and assisting dermatologists in complex cases, though further research is necessary to optimize their integration into clinical practice.
- IFU claims interpretation:
- Intended purpose: the device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: the device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
- Indications: the device is indicated for use on images of visible skin structure abnormalities for supporting the assessment of all skin diseases listed and described in the ICD-11 (code 14).
- Operating principle: The device is a computational medical tool leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures.
- Part of the body: the device is intended to analyse images of the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis). Due to the nature of the device (stand-alone software), it doesn't come into contact with tissue or bodily fluids.
Article ID 86
- Title / Year: Systematic review of machine learning for diagnosis and prognosis in dermatology (2019)
- Author: Thomsen, K.
- DOI / PMID: 10.1080/09546634.2019.1682500
- Brief summary of the article: The article discusses the application of artificial intelligence (AI) and machine learning in dermatology, particularly focusing on the assessment of skin conditions through image analysis. With around 20% of patients consulting general practitioners for skin-related issues, the need for accurate and timely dermatological diagnoses is critical. This is underscored by the challenges posed by a limited number of dermatologists, especially in countries where early diagnosis can significantly reduce morbidity and mortality associated with conditions like actinic keratosis (AK) and melanoma (MM). The advancements in AI, particularly machine learning algorithms like convolutional neural networks (CNNs), have revolutionized image classification and diagnosis in dermatology. CNNs, inspired by the human visual cortex, process images through multiple layers, extracting features and enabling the software to distinguish between malignant and benign conditions. This technology has shown promise, with studies reporting high accuracy rates in detecting MM and differentiating non-melanoma skin cancers. The article emphasizes that AI tools can assist general practitioners by streamlining the diagnostic process, thus addressing logistical challenges and enhancing early detection. By analyzing large datasets of dermatological images, these AI systems can improve diagnostic accuracy and potentially outperform human specialists in some scenarios.
- IFU claims interpretation:
- Intended purpose: the device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: the device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
- Operating principle: The device is a computational medical tool leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures.
- Part of the body: the device is intended to analyse images of the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis). Due to the nature of the device (stand-alone software), it doesn't come into contact with tissue or bodily fluids.
Article ID 109
- Title / Year: Skin Lesion Analysis towards Melanoma Detection Using Deep Learning Network (2018)
- Author: Li, Y.
- DOI / PMID: 10.3390/s18020556
- Brief summary of the article: Recent advancements in software and artificial intelligence (AI) for assessing skin conditions have focused on improving the automatic recognition of melanoma, which accounts for a significant portion of skin cancer deaths. Early detection is crucial for increasing survival rates, but manual identification requires highly trained specialists and can suffer from variability among observers. To address these challenges, automated systems leveraging dermoscopy—a noninvasive imaging technique that enhances the visibility of skin lesions—are being developed. Dermoscopy images face several obstacles in automated melanoma detection, such as low contrast between lesions and surrounding skin, high visual similarity between melanoma and non-melanoma lesions, and variations in skin characteristics among patients. Effective skin lesion segmentation is a vital step in classification approaches, and recent studies have produced various algorithms for this purpose. Notably, deep learning techniques have emerged as powerful tools for both segmentation and classification tasks in melanoma detection. For instance, several studies have explored different models, including convolutional neural networks (CNNs), to improve segmentation accuracy and lesion classification. In this context, two key frameworks have been introduced: the Lesion Indexing Network (LIN) and the Lesion Feature Network (LFN). The LIN combines multi-scale fully convolutional residual networks with a Lesion Index Calculation Unit (LICU), which refines segmentation and classification tasks by assessing pixel importance. This dual-task approach has demonstrated improved performance in both lesion segmentation and classification. The LFN, designed for dermoscopic feature extraction, employs CNNs to analyze skin lesions effectively. The framework addresses the challenges of imbalanced datasets through techniques such as weighted softmax loss and batch normalization, enhancing overall performance. Comparative analyses show that LIN and LFN outperform existing models in lesion segmentation and dermoscopic feature extraction tasks, indicating a promising direction for future research in automatic melanoma detection systems.
- IFU claims interpretation:
- Intended purpose: the device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: the device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
- Part of the body: the device is intended to analyse images of the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis). Due to the nature of the device (stand-alone software), it doesn't come into contact with tissue or bodily fluids.
Guidance documents
- Title / Year: The WHO ICD-11 Classification of Dermatological Diseases: a new comprehensive online skin disease taxonomy designed by and for dermatologists (2021)
- Author: British Journal of Dermatology
- DOI / PMID: 10.1111/bjd.20656
- Brief summary of the article: The document discusses the adoption of the 11th Revision of the International Classification of Diseases (ICD-11) by the World Health Assembly in May 2019, addressing the limitations of the previous ICD-10, which was released three decades ago. It emphasizes the significant input from dermatologists during the development process, led by the World Health Organization (WHO) since 2007, with the International League of Dermatological Societies (ILDS) actively involved. Key features of ICD-11 include:
- Integration and Adaptability: It allows for a more logical framework for skin disorders and better adaptability for healthcare research and management.
- Expanded Representation: The "Diseases of the skin" chapter has been restructured to include over 2000 dermatological entities, making rare and novel skin diseases individually represented.
- Enhanced Coding and Data Analysis: ICD-11 can link directly to patient record systems, improving diagnosis recording and data analysis.
- New Classifications: A dermatology-specific classification, the ICD-11 Classification of Dermatological Diseases (ICDD), has been created, which integrates with the ICD-11 electronic platform.
- Global Impact: The revisions increase the visibility of skin diseases and highlight their impact on health and well-being. The document concludes by encouraging the dermatology community to continue refining the classifications, ensuring their ongoing relevance and effectiveness, and notes that all current versions of ICD-11 are available online for public access.
- IFU claims interpretation:
- Intended purpose: the device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: the device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
- Part of the body: the device is intended to analyse images of the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis). Due to the nature of the device (stand-alone software), it doesn't come into contact with tissue or bodily fluids.
Description of the medical condition
Skin conditions such as melanoma, basal cell carcinoma (BCC), and acne vulgaris pose significant challenges in early detection and management. Advanced technologies, particularly artificial intelligence (AI) and machine learning (ML), are transforming dermatological diagnostics.
Melanoma is a particularly aggressive form of skin cancer whose incidence has been rising globally. Early detection is critical for improving survival rates. Traditional diagnostic methods, including visual inspections and histopathological examinations, often delay treatment. AI-driven tools are being developed to enhance diagnostic accuracy. For instance, the Dermalyser® app demonstrated a sensitivity of 95.2% and specificity of 60.3% in distinguishing melanoma from benign lesions, achieving an AUROC of 0.960 (1). Similarly, the ViT-GradCAM model showed a classification accuracy of approximately 96.6% across seven skin lesion categories, providing a user-friendly interface for real-time diagnosis (2). Other studies emphasize the use of deep learning algorithms trained on comprehensive datasets, like the HAM10000, which enhances diagnostic performance, achieving up to 84.3% accuracy (3). Notably, a mobile health app, SkinVision, achieved an overall sensitivity of 86.9% in detecting skin lesions (4), highlighting the integration of AI in clinical practice.
Acne vulgaris, prevalent among adolescents and adults, significantly impacts quality of life. Traditional assessment methods can be subjective and time-consuming. The AcneGrader framework introduces an ensemble of deep learning models for grading acne severity, aiming for greater accuracy (5). This innovation leverages AI's ability to classify lesions based on precise criteria, potentially improving treatment outcomes.
Recent studies demonstrate that AI tools not only streamline the diagnostic process but also offer insights that may be challenging for human practitioners to discern. By incorporating algorithms that analyze skin lesions captured through smartphone cameras, healthcare providers can facilitate timely diagnosis and enhance patient management (6). These developments underscore the transformative potential of AI in dermatology, making dermatological care more accessible and efficient.
WHO ICD-11 Classification of Dermatological Diseases
The WHO ICD-11 Classification of Dermatological Diseases represents a significant advancement in the classification and management of skin disorders, addressing the limitations of the previous ICD-10. Developed with substantial input from dermatologists and organizations like the International League of Dermatological Societies, ICD-11 introduces a more logical framework that enhances the adaptability of healthcare research and management. The Diseases of the skin
chapter now encompasses over 2000 dermatological entities, ensuring that even rare and novel conditions are accurately represented.
Key features include improved coding capabilities that facilitate integration with patient record systems, thereby enhancing diagnosis recording and data analysis. The introduction of the ICD-11 Classification of Dermatological Diseases (ICD) provides a dermatology-specific classification that aligns with the broader ICD-11 electronic platform. This comprehensive update not only increases the visibility of skin diseases but also emphasizes their significant impact on health and well-being.
In conclusion, the document encourages the dermatology community to continue refining these classifications to maintain their relevance and effectiveness, while all versions of ICD-11 remain accessible online for public use. This initiative marks a critical step toward improving the understanding and management of dermatological diseases globally (7).
Different approaches to dermatological conditions detection
Skin cancer, particularly melanoma, is a major global health concern, with increasing incidence rates largely attributed to exposure to ultraviolet (UV) radiation. Early detection and treatment are critical for improving patient outcomes, yet differentiating malignant lesions from benign ones poses a significant challenge for both primary care physicians (PCPs) and dermatologists (1,3). Traditional diagnostic methods, including visual examinations and histopathological assessments, can be time-consuming and often require specialized expertise, potentially leading to delays in treatment (6).
Recent advancements in AI and machine learning have revolutionized skin cancer diagnostics. Various AI-driven tools have been developed to assist healthcare providers in the rapid and accurate identification of skin lesions using smartphone technology and other imaging modalities (2,8). For instance, tools like Dermalyser® have shown promising results, achieving high sensitivity (95.2%) and specificity (60.3%) in distinguishing melanoma from benign lesions in a real-world primary care setting (1). Similarly, the SkinVision app demonstrated an overall sensitivity of 86.9% and specificity of 70.4%, highlighting the potential of mobile health applications in improving skin cancer screening accessibility (4).
The ViT-GradCAM model represents another innovative approach that utilizes deep learning techniques to enhance the classification of skin lesions. This model combines data augmentation and synthetic data generation to address challenges associated with limited datasets, achieving a classification accuracy of 96.6% across various skin lesion types (2). Machine learning models trained on comprehensive datasets like the HAM10000 have shown superior performance in classifying skin lesions compared to traditional methods, underscoring the importance of robust training datasets in developing effective diagnostic tools (3).
Moreover, the integration of AI in teledermatology has become increasingly relevant, especially during the COVID-19 pandemic, which exacerbated access issues to dermatological care (6). AI frameworks designed to analyze images captured by consumer-grade smartphones have emerged as viable solutions, providing timely evaluations of skin lesions and enabling remote consultations (9,10). These advancements not only facilitate faster diagnosis but also enhance patient management by allowing PCPs to make informed decisions based on AI-generated insights (8).
Recent advancements in artificial intelligence (AI) and deep learning have significantly improved the detection and classification of skin lesions, particularly melanoma, which is crucial for enhancing early diagnosis and treatment outcomes. Maqsood et al. (2023) highlight the development of automated detection methods that integrate image acquisition, feature extraction, segmentation, and classification, utilizing convolutional neural networks (CNNs) to differentiate between benign and malignant lesions in dermoscopic images. This integration addresses the limitations of traditional skin examination methods, which can be time-consuming and prone to human error (11).
Thomsen (2019) emphasizes the growing need for accurate dermatological diagnoses, especially in regions with limited access to dermatologists. Machine learning, particularly through CNNs, has revolutionized image analysis in dermatology, allowing for high accuracy rates in detecting melanoma and non-melanoma skin cancers. These AI tools can streamline the diagnostic process, thereby enhancing early detection and potentially outperforming human specialists in some cases (12).
Li et al. (2018) further explore the challenges faced in automated melanoma detection, such as low contrast in dermoscopic images and variability among observers. They present frameworks like the Lesion Indexing Network (LIN) and Lesion Feature Network (LFN) that leverage deep learning for improved segmentation and classification accuracy. These innovations indicate a promising future for AI-driven systems in dermatology, aimed at supporting healthcare professionals in making informed clinical decisions while addressing the critical issue of skin cancer early detection (13).
While the development of these AI-based diagnostic tools holds great promise, it is essential to continue validating their performance through prospective studies to ensure their clinical utility and integration into routine dermatological practice (3,4). Ongoing research and innovation in this field aim to further improve the accuracy and efficiency of skin cancer detection, ultimately contributing to better patient outcomes and enhanced healthcare delivery systems.
Possible risks of the use of software and artificial intelligence in dermatological diagnostics
The use of software and artificial intelligence (AI) in dermatological diagnostics introduces several potential risks, as highlighted in various articles:
- Diagnostic Inaccuracy: While AI tools can enhance diagnostic accuracy, there is a risk of misclassification of skin lesions. False positives may lead to unnecessary anxiety and invasive procedures, while false negatives could delay critical treatment for malignant conditions, particularly melanoma (11,12).
- Data Quality and Bias: The performance of AI models heavily relies on the quality and representativeness of the training datasets. Limited or biased datasets may result in AI systems that are less effective for diverse populations, potentially exacerbating health disparities (13).
- Overreliance on Technology: As healthcare providers increasingly utilize AI for decision-making, there is a risk of overreliance on technology. This may lead to diminished clinical judgment, where practitioners might not thoroughly evaluate lesions without the assistance of AI tools (12).
- Privacy and Security Concerns: The integration of AI in dermatological diagnostics often involves processing sensitive patient data. There is a risk of breaches in privacy and data security, particularly if proper safeguards are not implemented (11).
- Implementation Challenges: The transition to AI-assisted diagnostics may encounter resistance from healthcare providers accustomed to traditional methods. Additionally, there may be challenges in integrating AI tools into existing workflows, which could hinder their effectiveness (12).
- Limited Clinical Validation: Many AI tools have not undergone extensive clinical validation through prospective studies. The lack of rigorous testing raises concerns about their reliability and effectiveness in real-world settings (11,13).
- Regulatory Compliance: The rapid development of AI technologies may outpace regulatory frameworks, leading to challenges in ensuring that AI tools meet safety and efficacy standards. Inadequate regulation may result in the use of subpar or unsafe diagnostic tools (12).
By addressing these risks through careful validation, robust datasets, and ongoing clinician training, the potential of AI in dermatological diagnostics can be harnessed while safeguarding patient outcomes and ensuring effective healthcare delivery.
Minimization and management of possible risks
Based on the articles reviewed, the methods for minimizing and managing risks related to the use of software and AI in dermatological diagnostics include:
- Enhanced Algorithm Development: The studies emphasize the importance of developing sophisticated algorithms that integrate deep learning techniques, such as convolutional neural networks (CNNs), which can improve accuracy in differentiating between malignant and benign lesions. For example, Maqsood et al. (2023) focus on frameworks that combine image acquisition, feature extraction, segmentation, and classification, which streamline the diagnostic process and reduce the likelihood of misdiagnosis (11).
- Comprehensive Training Datasets: The need for extensive and diverse datasets is highlighted in the literature. Li et al. (2018) point out that machine learning models trained on robust datasets, like HAM10000, demonstrate superior performance in classifying skin lesions. Ensuring that training datasets are representative of various skin types and conditions can mitigate risks associated with algorithmic bias and improve diagnostic accuracy (13).
- Real-World Validation: Thomsen (2019) discusses the necessity of validating AI tools in real-world settings. This ongoing evaluation helps to confirm their clinical utility and performance, ensuring that they remain effective across different populations and environments (12).
- Integration with Clinical Practices: The articles suggest that AI tools should complement traditional diagnostic methods rather than replace them. By integrating AI insights with the clinical expertise of healthcare providers, the risk of misdiagnosis can be minimized. Clinicians can use AI-generated insights to support their decision-making process, ensuring a comprehensive evaluation of skin lesions (11,12).
- User Training and Education: The importance of training healthcare professionals on the use of AI-driven tools is highlighted. Ensuring that users are well-versed in interpreting AI outputs and understanding their limitations can prevent overreliance on technology and improve the overall quality of care (12).
- Feedback Mechanisms: The studies advocate for implementing feedback mechanisms that allow healthcare providers to report their experiences with AI tools. Continuous feedback can inform iterative improvements, ensuring that AI systems evolve to meet clinical needs and address any emerging risks (11).
- Robust Data Security Protocols: Although not always explicitly mentioned, the need for stringent data security measures is implied in the context of protecting patient information and maintaining confidentiality when using AI technologies (12).
By applying these methods, as highlighted in the articles, healthcare providers can effectively manage risks associated with the implementation of AI and software in dermatological diagnostics, thereby enhancing patient outcomes and the reliability of the diagnostic process.
Identification of similar devices
Different similar devices to the device which are certified devices have been identified (see table below).
Name | Website | Country of Origin | How to Access and Use | Founding Year | Value Proposition | Targeted Medical Conditions | CE Marking | Medical Device Risk Classification under MDR | Hardware Requirement |
---|---|---|---|---|---|---|---|---|---|
SkinVision | skinvision.com | The Netherlands | Download in App Store or Browser | Not Found | Skin cancer risk assessment via photos, recommendations for next steps | Skin Cancer | Yes | Not Found | Mobile devices with certain camera quality |
DermEngine | dermengine.com | Canada | Web, iOS, Android, tvOS | 2014 or 2012 | Imaging, documentation, analysis of skin, hair, nails with AI support | Dermatology-related conditions | Not Found | Not Found | Compatible with various devices & integration with dermoscopes |
Triage | triage.com | Canada | Not Found | Not Found | Screening for 588 skin disorders using medical-grade AI | Skin Disorders | Not Found | Not Found | Not Found |
First Derm | firstderm.com | USA | Web, iOS, Android app | 2014 | Online dermatology consultations, not a substitute for doctor's visit but offers medical guidance | Skin Conditions | Not Found | Not Found | Smartphone, optional use of HÜD device for mole scanning |
Cureskin | cureskin.com | India | Web, iOS, Android app | 2016 or 2017 | AI-powered analysis and expert dermatologist consultations for skin, hair, body, and personal care concerns | Skin and Hair Conditions | Not Found | Not Found | Smartphone |
MoleMapper | molemapper.org | [Not Found] | iOS App, Android App | Not Found | Mapping, measuring, and monitoring moles to detect potentially harmful changes, aiding in melanoma research | Melanoma Detection | Not Found | Not Found | Smartphone with camera |
Eczema Tracker | eczematracker.com | [Not Found] | iOS App, Android App | 2016 | Analyzing, managing, and controlling eczema; tracking and analyzing triggers; getting advice; maintaining medicine usage record; staying updated on eczema-related news | Eczema | Not Found | Not Found | Smartphone |
VisualDx | visualdx.com | USA (Rochester, NY) | PC, iPad/iPhone, Android Tablet/Phone | 2001 | Clinical decision support system aiding in differential diagnosis via peer-reviewed images, intended for medical practitioners including primary care practitioners | Various Medical Conditions | Not Found | Not Found | PC, Tablet, or Smartphone |
MoleScope (by Fotofinder) | molescope.com | Canada (Vancouver, BC)4 | Attach to specified smartphones or tablets | 2012 | Imaging, archiving, and communication of skin conditions | Mole imaging, other skin conditions like acne, eczema, psoriasis | Yes | Not Found | Specified iOS and Android devices |
Conclusion on the comparison with similar devices
A comparison of the main technical, clinical and biological characteristics for the device and the similar devices: SkinVision and MoleScope is provided in the table below. The instructions for use of these devices have been used in order to fill in the table.
In comparing the device with CE-marked devices SkinVision and MoleScope, key differences emerge in technical and clinical characteristics, including intended use, target users, and biological applications.
- Technical Characteristics: the device is a computational, software-only device that analyzes images of skin structures without direct patient contact. SkinVision is an over-the-counter (OTC) mobile app for layperson use, providing risk assessments for skin cancer via smartphone images. MoleScope is a mobile dermatoscope requiring physical contact with the skin to capture high-resolution images, primarily targeting healthcare professionals and laypersons. Legit.Health, however, stands out by covering a wider range of skin structures, including the epidermis, dermis, and cutaneous vasculature, with a focus on aiding healthcare practitioners.
- Clinical Characteristics: The intended use of the device is to support healthcare practitioners by quantifying visible clinical signs and suggesting potential International Classification of Diseases (ICD) classifications, which enhances efficiency in clinical assessments. In contrast, SkinVision is geared toward self-assessment, guiding lay users on skin lesion monitoring and suggesting professional follow-up when needed. MoleScope, primarily used in clinical settings, allows for detailed image capture and documentation but does not provide automated assessments.
- Warnings, Contraindications, and Patient Safety: Each device includes specific contraindications for use in certain anatomical locations and skin conditions. Our device, as software-only, does not directly interact with bodily structures, which reduces physical risks. SkinVision and MoleScope warn users against imaging lesions in certain body areas (e.g., those covered with hair or near mucosal surfaces), noting their devices' limitations in capturing or processing images accurately.
Overall, our device diverges significantly by focusing on medical professionals, offering detailed structural analysis and avoiding the direct patient-device interface. SkinVision and MoleScope both prioritize ease of use for broader audiences, with MoleScope emphasizing clinical image quality and SkinVision focusing on user-guided skin health management.
The information from articles on MoleScope and SkinVision will be incorporated into the state of the art (SOTA) documentation, as these devices offer similar functionalities and intended purposes. Comparing the technical and clinical characteristics with MoleScope and SkinVision provides a broader perspective on the capabilities, challenges, and advancements within the field of AI-powered dermatology tools. This approach will strengthen the contextual understanding of our position within the current technological landscape of dermatologic imaging and AI analysis, supporting its development and regulatory documentation.
Technical characteristics
Legit.Health Plus | SkinVision | MoleScope | Similar (Y/N/Partial) | |
---|---|---|---|---|
Manufacturer | AI Labs Group S.L. | SkinVision B.V. | Dermengine | |
Description | The device is computational software-only medical device leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures. Its principal function is to provide a wide range of clinical data from the analyzed images to assist healthcare practitioners in their clinical evaluations and allow healthcare provider organisations to gather data and improve their workflows. The generated data is intended to aid healthcare practitioners and organizations in their clinical decision-making process, thus enhancing the efficiency and accuracy of care delivery. The device should never be used to confirm a clinical diagnosis. On the contrary, its result is one element of the overall clinical assessment. Indeed, the device is designed to be used when a healthcare practitioner chooses to obtain additional information to consider a decision. | At the centre of the service is the SkinVision app, which is a regulated medical device that merges AI technology with the expertise of skin health professionals and dermatologists. SkinVision is a service of choice whether you want to address your most immediate concerns, learn what steps you should take next, understand your skin risk profile and introduce the most intelligent skin health regimen to your seasonal rhythm. | MoleScope is a mobile dermatoscope that allows you to capture a high-resolution, detailed view of the skin through magnification and specialized lighting. The images are accessible anywhere at anytime through the secure platform for analysis and diagnosis. The device is indicated for use in combination with a smartphone or a tablet when magnified images of the skin, scalp, hair, nail, and other body surfaces of any patient demographic are required. | Partial |
Principle of operation | The device is a computational medical tool leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures. | Use the SkinVision Smart Check Camera to check your skin and get instant risk assessments. Create your Body Map to track your skin spots regularly for changes over time. Identify your Skin Type and Risk Profile to learn the most effective ways to protect yourself from skin cancer risk and receive personalised advice, assessment and care. | DermEngine's analytical tools allow you to examine and compare images in-depth on your tablet and desktop web browsers. Advanced imaging and analytic tools provide state of the art features for dermatology. Designed to work seamlessly with the MoleScope dermoscopy attachment, the DermEngine app provides precise and efficient documentation for your busy clinical workflow. The comprehensive mobile medical imaging solution syncs with the DermEngine web platform for image archiving and patient management. | No |
Biological characteristics
Legit.Health Plus | SkinVision | MoleScope | Similar (Y/N/Partial) | |
---|---|---|---|---|
Components | The device is computational software-only medical device leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures. | SkinVision App | Device Models A / A+ / B / B+: A mobile dermatoscope device with a contact cap and a rubber attachment, a non-contact cap, a microfiber cleaning cloth, a microfiber device bag and a charging cable. Device Model U: A mobile dermatoscope device with a contact cap, a universal attachment plate, a non-contact cap, a microfiber cleaning cloth, a microfiber device bag and a charging cable. | No |
Parts of human body in contact with product | The device is intended to analyse images of the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis). Due to the nature of the device (stand-alone software), it doesn't come into contact with tissue or bodily fluids. | Not applicable | Not applicable | Yes |
Lifetime | The expected operational lifetime of the device is established at 5 years, which is subject to regular software updates and the lifecycle of the integrated components and platforms. The lifetime will be increase in equivalent spans as the design and development continues and maintenance and re-design activities are carried out. This timeline accounts for the expected evolution of the underlying operating systems and tools, the progression of medical device technology, and the necessary update cycles to maintain security and operability. | Information not provided in manufacturer's IFU | Information not provided in manufacturer's IFU | Unknown |
Clinical characteristics
Legit.Health Plus | SkinVision | MoleScope | Similar (Y/N/Partial) | |
---|---|---|---|---|
Intended use | The device is a computational software-only medical device intended to support Health Care Practitioners in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs interpretative distribution representation of possible International Classification of Diseases (ICD) categories. | The SkinVision Service is a software-only, over-the-counter (OTC), mobile medical application, which is intended for use on consumer mobile devices by laypeople. The SkinVision Service consists of an assessment module that generates an immediate indication of risk for the most common types of skin cancer based on a picture of the skin lesion of concern, taken with the app on the mobile device. Based on the assessment, the SkinVision Service provides a recommendation whether to continue monitoring or recommends to visit a health professional for a further review of the skin lesion. The application also facilitates keeping track of skin lesions and provides information on the photographed lesions that may be used when seeking professional healthcare advice. The SkinVision Service augments already existing self-assessment techniques of skin lesions and is not an alternative to healthcare professionals. The SkinVision Service is not intended for use on persons under the age of 18 years old. The SkinVision Service does not diagnose skin cancer, does not provide any other diagnosis. | MoleScope is a battery-powered dermatoscope intended to be attached to smartphones and tablets to take images of skin, scalp, hair, nail, and other body surfaces with high magnification and cross-polarized light. It is a reusable, non-invasive device for imaging and documentation only. | Partial |
Intended patient population | The device is intended for use on images of skin from patients presenting visible skin structure abnormalities, across all age groups, skin types, and demographics. | Laypeople. | Laypeople. | Partial |
Target user | Healthcare professionals + IT professionals | It is intended for use on consumer mobile devices by laypeople. | The device is intended to be used by professional healthcare workers or lay persons. | No |
Warnings, cautions and precautions | In case of observing an incorrect operation of the device, notify us as soon as possible. You can use the email support@legit.health. We, as manufacturers, will proceed accordingly. Any serious incident should be reported to us, as well as to the national competent authority of the country. | 1. Do not install or use the SkinVision App on a device with non-original iOS or Android ('jailbroken' or 'rooted') software. 2. Before using the SkinVision App, check that the camera lens is clean and not obstructed by anything. Wipe the lens carefully with a soft cloth if it is dirty. 3. Do not cover the flashlight source or the camera in any way while the photos are being captured. 4. Do not upload or transmit content of any type that may infringe or violate the rights of any party. 5. Do not disable, modify, "hack" or otherwise interfere with the proper functioning of this software. 6. Due to the functionality of the algorithm, in certain cases you may receive different risk assessment outcomes for different photos taken in rapid succession of the same skin spot. In such cases, err on the side of caution and always visit a healthcare professional. | • Do not look directly into the bright LED light. Eyes must be closed during facial examination. • Do not attempt to open the device for any reason. • Do not attempt to change the device battery. The battery is not replaceable. • Do not use the device to image sensitive areas, open wounds, and surfaces near the eyes, ears, nose, and mouth. • Check your device before use if it has been dropped. • Follow the instructions in this document to clean and disinfect contact and non- contact caps in between patients. • Keep the device and detachable components out of reach of children and pets. If swallowed, seek medical attention immediately. • Do not use the device while it is in charging mode. • This device contains magnets. Do not use it without first consulting your doctor if you wear a pacemaker or other medical implant. | No |
Contraindications | We advise not to use the device in: Skin structures located at a distance greater than 1 cm from the eye, beyond the optimal range for examination. Skin areas that are obscured from view, situated within skin folds or concealed in other manners, making them inaccessible for camera examination. Regions of the skin showcasing scars or fibrosis, indicative of past injuries or trauma. Skin structures exhibiting extensive damage, characterized by severe ulcerations or active bleeding. Skin structures contaminated with foreign substances, including but not limited to tattoos and creams. Skin structures situated at anatomically special sites, such as underneath the nails, requiring special attention. Portions of skin that are densely covered with hair, potentially obstructing the view and hindering examination. | It is not advised to use the app to take a photo of a skin spot or lesion that: 1. Is close in color with the surrounding skin, for example, a skin spot on (very) dark skin, a white patch on fair skin, or on sunburnt skin, 2. Is on a darker skin type (IV, V and VI), due to camera limitations, 3. is surrounded by multiple skin spots (for example: a rash or an irritation on the skin's surface. It can look like a cluster of small red bumps, blotches, or reddened areas that may cause itching or burning.) 4. is surrounded by non-intact skin (e.g. open sores, ulcers, bleeding, scabs), 5. is under the nail, 6. is close to a (visible) area with scar(s), 7. contains foreign matter (e.g. marker, tattoo, sunscreen, skin cream, powder, etc.), 8. is covered by a significant amount of hair, 9. is on mucosal surfaces (e.g. lips, genitals), 10. is on or near a skinfold (e.g eyelid, navel), 11. is not on human skin. | The device should not be used on open wounds or on any other sensitive areas. | No |
CE mark | No | Yes | Yes | Partial |
Valid clinical association
Demonstration of valid clinical association had the objective to demonstrate the association of MDSW output (intensity, count and extent of visible clinical signs) and the organs affected as well as the association oft he visible clinical signs measured and the interpretative distribution representation of possible ICD categories. Fort hat purpose, different searches have been performed in Section 5.1 and Section 5.2. The demonstration of valid clinical association should cover:
- Has the quantification of intensity, count, and extent of visible clinical signs related to the epidermis, its appendages (including hair, hair follicles, sebaceous glands, apocrine and eccrine sweat gland apparatus, and nails), as well as associated mucous membranes (such as conjunctival, oral, and genital), the dermis, cutaneous vasculature, and subcutaneous tissue (subcutis) been validated against the International Classification of Diseases (ICD) categories?
- Are the visible clinical signs detected validated as appropriate for the interpretative distribution representation of possible ICD categories?
The answer to these questions is provided in the Conclusions on Valid Clinical Association
section.
Valid clinical association of the visible skin structures abnormalities
The device provides quantifiable data on the intensity, count and extent of visible clinical signs such as erythema, desquamation, and induration, among others. A systematic literature search has been performed in order to demonstrate the valid clinical association between the areas or organs affected and the visible clinical signs measured.
Identification of data (Stage 1) - Literature search
In order to establish the correlation between the MDSW output (intensity, count and extent of visible clinical signs) and the body parts affected, a systematic literature search has been performed.
- Date of the search: 2024-10-14
- Period covered by the search: 2014-10-14 - 2024-10-14
- Sources used: PubMed.gov. PubMed is a free resource supporting the search and retrieval of biomedical and life sciences literature. The database citations (more than 34 million citations and abstracts of biomedical literature) primarily stem from the biomedicine and health fields, and related disciplines such as life sciences, behavioral sciences, chemical sciences, and bioengineering.
- Search algorithms: ("skin structure" OR "epidermis" OR "dermis" OR "hair" OR "hair follicle" OR "sebaceous gland*" OR "nails") AND ("visible clinical sign*" OR "skin condition" OR "erythema" OR "desquamation" OR "induration" OR "dryness" OR "swelling") AND ("intensity" OR "count" OR "quantif*")
- Language: English
- Filters: Full text, abstract human
Screening and appraisal (Stages 2 and 3)
As a result of the search, 102 articles were retrieved.
The 102 articles are stored in the dedicated folder of the technical documentation (Literature Search Records
).
97 articles were discarded as they were not related to the device under evaluation or the medical condition. The remaining 5 articles were appraised according to the criteria for appraisal defined in the Clinical Evaluation Plan:
- Correlation between MDSW output and condition (Criteria 1)
- Article does not describe the correlation between the clinical visible signs and the measured skin structures (Score: 1)
- Article has a limited description of the relation between the condition and MDSW output (Score: 2)
- Article describes the correlation between the condition and the MDSW output and it is well explained and developed (Score: 3)
- Methodology (Criteria 2)
- Methods not described or described very poorly (Score: 1)
- Limited description of the methodology applied (Score: 2)
- Methods are well described (Score: 3)
- Medical condition (Criteria 3)
- Referred to other conditions (non-visible skin conditions) (Score: 1)
- Different type of clinical conditions observed in visible skin (Score: 2)
- Type of visible clinical skin conditions under evaluation (Score: 3)
Finally, 4 articles from PubMed were selected for their evaluation as part of the valid clinical association. A summary of the whole process in described in the following flowchart.
The articles selected for their evaluation are summarized in the following table.
Reference | Name article | Link |
---|---|---|
001 | Long-term efficacy and safety of nonablative monopolar radiofrequency in the treatment of moderate to severe acne vulgaris. | https://doi.org/10.1002/lsm.23757 |
014 | Nipple Skin Trauma in Breastfeeding Women During Postpartum Week One. | https://doi.org/10.1089/bfm.2017.0217 |
017 | Increased number of mast cells in the dermis in actinic keratosis lesions effectively treated with imiquimod. | https://doi.org/10.1111/1346-8138.13821 |
030 | Development of a new classification and scoring system for scalp conditions: Scalp Photographic Index (SPI). | https://doi.org/10.1080/09546634.2023.2181655 |
The summary for the appraisal is summarized below.
ID | Title | DOI | Criteria 1 | Criteria 2 | Criteria 3 | Punctuation | Final Appraisal |
---|---|---|---|---|---|---|---|
1 | Long-term efficacy and safety of nonablative monopolar radiofrequency in the treatment of moderate to severe acne vulgaris. | 10.1002/lsm.23757 | 2 | 3 | 3 | 8 | Yes |
14 | Nipple Skin Trauma in Breastfeeding Women During Postpartum Week One. | 10.1089/bfm.2017.0217 | 3 | 3 | 3 | 9 | Yes |
17 | Increased number of mast cells in the dermis in actinic keratosis lesions effectively treated with imiquimod. | 10.1111/1346-8138.13821 | 2 | 3 | 3 | 8 | Yes |
30 | Development of a new classification and scoring system for scalp conditions: Scalp Photographic Index (SPI). | 10.1080/09546634.2023.2181655 | 3 | 3 | 3 | 9 | Yes |
43 | Molecular characterization of xerosis cutis: A systematic review. | 10.1371/journal.pone.0261253 | 1 | 2 | 2 | 5 | No |
Analysis of the data
The valid clinical association analysis aims to collectively evaluate all the appraised information, in terms of weight and significance. The present section summarizes the main conclusions extracted from the analysis of the articles and guidelines including conclusions supporting a valid clinical association between MDSW output and the condition and thus conformance with the applicable GSPRs.
Article ID 1
- Title / Year: Long-term efficacy and safety of nonablative monopolar radiofrequency in the treatment of moderate to severe acne vulgaris (2024).
- Author: Manuskiatti, W.
- DOI / PMID: 10.1002/lsm.23757
- Brief Summary of the article: The article on acne vulgaris (AV) discusses the pathophysiology and clinical manifestations of the condition, highlighting key factors such as excess sebum production, follicular hyperkeratinization, inflammation, and Cutibacterium acnes colonization. These processes result in visible skin changes, including comedones, papules, pustules, and nodules, which are characteristic of AV. Analyzing these visible indicators is crucial for clinical assessments and management of the condition. Advanced tools that analyze skin images can provide detailed insights into these dermatological manifestations, aiding healthcare practitioners by offering additional data that can be integrated into their clinical evaluations. This allows for more informed decision-making and supports the monitoring of treatment efficacy over time. The clinical association between the observed skin abnormalities and the underlying pathophysiological mechanisms of AV further underscores the importance of accurate visual analysis in the diagnostic process.
- Relevance for Legit.Health Valid Clinical Association:
- Clinical and Pathophysiological Insights: The article offers valuable details on the pathophysiology and clinical manifestations of acne vulgaris, covering factors such as sebum production, follicular hyperkeratinization, inflammation, and Cutibacterium acnes colonization. This aligns with Legit.Health's focus on analyzing visible indicators, as understanding these processes is essential for accurate image-based analysis of skin conditions.
- Observed Clinical Signs: The article emphasizes the significance of visible clinical signs (e.g., comedones, papules, pustules, nodules) in evaluating acne severity. This supports the device's objective of quantifying and counting such signs, which aids in tracking disease progression and treatment efficacy over time.
- Utility of Image-Based Analysis: Advanced image analysis tools, such as those Legit.Health employs, can enhance dermatological assessments by offering precise measurements of clinical indicators. This article further validates the need for image-based recognition, which can capture and quantify subtle changes in skin features, critical for assessing acne and similar dermatological conditions.
- Long-term Treatment Monitoring: Given the article's focus on the treatment efficacy and safety over time, it reinforces the value of a tool like Legit.Health, which supports longitudinal data collection and tracking. This is beneficial for both healthcare providers and patients in monitoring skin condition improvements or flare-ups.
Article ID 14
- Title / Year: Nipple Skin Trauma in Breastfeeding Women During Postpartum Week One (2018).
- Author: Nakamura, M.
- DOI / PMID: 10.1089/bfm.2017.0217
- Brief Summary of the article: This study investigates nipple trauma, a common issue that affects breastfeeding mothers, especially in the first week postpartum. Nipple trauma leads to pain and discomfort, often causing mothers to stop breastfeeding. Various prevention and treatment methods have been studied, but there is no consensus on the most effective strategy due to a lack of standardized definitions and varying interpretations of nipple trauma severity. The study aimed to classify signs of nipple trauma and describe changes in nipple skin during the first postpartum week through a two-phase observational study involving 50 breastfeeding women. Photographs of the participants' nipples were taken daily, and changes such as erythema, swelling, blistering, fissures, and scabbing were analyzed. Erythema was the most frequent sign, while scabbing was associated with the highest pain scores. Multiple signs often appeared simultaneously. The second phase of the study assessed the reliability of these signs using midwife evaluations, confirming the consistency of the identified trauma indicators. The study emphasizes the need for standardized classifications to better understand nipple trauma and improve breastfeeding outcomes.
- Relevance for Legit.Health Valid Clinical Association:
- Quantifiable Clinical Signs: The study identifies visible signs such as erythema, swelling, blistering, fissures, and scabbing, which align with Legit.Health's emphasis on quantifying and categorizing visible clinical signs.
- Pain and Severity Correlation: The article connects scabbing with the highest pain scores, which reinforces the importance of capturing clinical signs linked to symptom severity. Legit.Health's system could leverage such correlations to provide healthcare providers with insights into patient discomfort based on visual indicators.
- Standardized Classifications: The study highlights the need for consistent classifications of trauma indicators, mirroring the device's function of offering standardized, image-based assessments. Legit.Health's algorithmic approach to interpreting skin abnormalities could contribute to more standardized evaluations across various healthcare settings.
- Reliability of observations: The article's focus on reliability aligns with Legit.Health's objective to provide consistent, quantifiable data that healthcare professionals can trust for patient assessment and treatment monitoring over time.
Article ID 17
- Title / Year: Increased number of mast cells in the dermis in actinic keratosis lesions effectively treated with imiquimod (2017).
- Author: Oyama, S.
- DOI / PMID: 10.1111/1346-8138.13821
- Brief Summary of the article: The article focuses on assessing clinical signs in actinic keratosis (AK) and their measurement during imiquimod treatment. AK is a skin cancer in situ commonly observed on sun-exposed areas, particularly in individuals with fair skin, older age, and chronic ultraviolet (UV) exposure. In terms of clinical assessment, AK lesions were examined using dermoscopy, focusing on erythema and a “strawberry” pattern characterized by telangiectatic vessels around hair follicles and keratotic plugs. Erythema was graded on a scale from 0 to 4, with 0 indicating no erythema and 4 indicating very severe erythema. The study observed that while the erythema score increased significantly in both responsive (cured) and unresponsive lesions after treatment, there was no significant difference between the two groups. However, dermatoscopic features such as the strawberry pattern, which was present in 76% of lesions before treatment, were significantly reduced after treatment, particularly in responsive lesions where the pattern dropped to 17%. In addition to clinical signs, the study also used histopathological analysis, measuring the number of Ki-67-positive proliferative cells in the epidermis and CD117-positive mast cells in the dermis. These findings are significant for monitoring the clinical and histopathological response of AK lesions to imiquimod therapy, providing valuable insights into the treatment's effectiveness.
- Relevance for Legit.Health Valid Clinical Association:
- Quantifiable Clinical Signs: The study examines erythema and a distinct "strawberry" pattern in actinic keratosis (AK) lesions, providing a grading scale (0-4) for erythema severity. Legit.Health quantifies erythema intensity, and recognizes dermatoscopic patterns in images, enhancing its ability to assess and quantify visible clinical signs.
- Pattern Recognition: The "strawberry" pattern around hair follicles, characterized by telangiectatic vessels and keratotic plugs, was significantly reduced in responsive lesions post-treatment. Legit.Health's image-based analysis can help detect and track similar patterns in skin images, supporting the assessment of treatment effectiveness for skin conditions.
- Correlation with Treatment Response: The study's observation that erythema scores increased post-treatment in both responsive and unresponsive lesions, while the "strawberry" pattern diminished primarily in responsive lesions, highlights the importance of differentiating clinical signs in treated versus untreated conditions. Legit.Health's algorithms could similarly help monitor changes in response to treatment.
- Histopathological Data: The article's use of histopathological markers (Ki-67-positive cells and CD117-positive mast cells) to assess cellular response may not directly translate to image-based software, but it underscores the importance of correlating visual clinical signs with underlying biological changes—a concept that can enhance Legit.Health's interpretative abilities for assessing skin abnormalities.
Article ID 30
- Title / Year: Development of a new classification and scoring system for scalp conditions: Scalp Photographic Index (SPI) (2023).
- Author: Kim, B.
- DOI / PMID: 10.1080/09546634.2023.2181655
- Brief Summary of the article: The article introduces the Scalp Photographic Index (SPI), a tool designed to objectively and reproducibly grade five key features of scalp conditions: dryness, oiliness, erythema, folliculitis, and dandruff. These features are graded on a four-point scale (0: none, 1: mild, 2: moderate, 3: severe) based on magnified photographs of five areas of the scalp (vertex, anterior, posterior, right, and left). Dryness is assessed by the presence of fine lines or scales, oiliness by sebum accumulation and shine, erythema by redness or telangiectasia, folliculitis by erythema or pustules around hair follicles, and dandruff by the size and visibility of scales. The total SPI score, which can range from 0 to 75, reflects the overall severity of scalp problems, and specific scalp types are defined by features scoring 2 or 3. This grading system was found to correlate significantly with both dermatologists' assessments and patients' self-reported symptoms, showing good internal consistency and high inter- and intra-rater reliability across various healthcare professionals.
- Relevance for Legit.Health Valid Clinical Association:
- Objective Measurement: The SPI utilizes a four-point grading scale to assess features such as dryness, oiliness, erythema, folliculitis, and dandruff. Similarly, Legit.Health quantifies visible clinical signs (e.g., erythema, desquamation) through measurable parameters such as intensity, count, and extent, enhancing the objectivity of skin assessments.
- Quantification and Representation: Just as the SPI generates a total score that reflects overall scalp severity, Legit.Health provides quantifiable data that aids in evaluating skin structure abnormalities. The data generated by Legit.Health can support healthcare practitioners in decision-making processes, much like the SPI informs assessments.
- Correlational Validity: The SPI's correlation with dermatologists' assessments and patients' self-reported symptoms underscores the importance of establishing clinical relevance. For Legit.Health, the ability to deliver an interpretative distribution representation of possible ICD categories suggests a potential for correlational validity with established medical classifications, aligning with the need for clinical relevance in diagnostic tools.
- Reliability: The article emphasizes high inter- and intra-rater reliability of the SPI across healthcare professionals. This reliability is crucial for Legit.Health, as the device is intended to assist a wide range of healthcare providers, ensuring that the quantifiable outputs are consistent and dependable for varied users.
- Focus on Visible Abnormalities: The SPI's assessment focuses on visible scalp features, which is parallel to the intended use of Legit.Health, where it analyzes images of visible skin structure abnormalities. This focus reinforces the relevance of the results produced by Legit.Health in clinical practice.
Valid clinical association of the International Classification of Diseases (ICD) categories
The device provides an interpretative distribution representation of possible International Classification of Diseases (ICD) categories. Several systematic literature searches have been performed in order to demonstrate the valid clinical association between the visible clinical signs measured and the most relevant and common ICD categories.
Identification of data (Stage 1) - Literature search
In order to establish the correlation between the MDSW output (interpretative distribution representation of possible International Classification of Diseases (ICD) categories) and the visible clinical signs measured, a systematic literature search has been performed.
- Source used: PubMed.gov. PubMed is a free resource supporting the search and retrieval of biomedical and life sciences literature. The database citations (more than 34 million citations and abstracts of biomedical literature) primarily stem from the biomedicine and health fields, and related disciplines such as life sciences, behavioral sciences, chemical sciences, and bioengineering.
- Language: English
- Filters: Full text, abstract, human
- Search: see table below
Search | Diseases | Search algorithm | Date of search | Search period |
---|---|---|---|---|
Search 1 | Precancerous conditions: actinic keratosis | ("actinic keratosis") AND ("erythema" OR "induration" OR "swelling" OR "desquamation" OR "nodule count" OR "depth"OR "visible clinical sign*" OR "skin condition") AND ("intensity" OR "count" OR "quantif*") | 2024-10-16 | 2014-10-16 / 2024-10-16 |
Search 2 | Malignant neoplasms and pigmented neoplasms | ("Malignant adnexal tumors" OR "Malignant epidermal neoplasms" OR "Malignant fibrous neoplasms" OR "Malignant lymphoproliferative conditions" OR "Malignant neoplasms" OR "Malignant vascular neoplasms" OR "Benign melanocytic neoplasms" OR "Disorders of pigmentation" OR "Malignant melanocytic neoplasms") AND ("erythema" OR "induration" OR "swelling" OR "desquamation" OR "nodule count" OR "visible clinical sign*" OR "skin condition") | 2024-10-18 | 2014-10-18 / 2024-10-18 |
Search 3 | Inflammatory conditions | ("Acne" OR "eczematous dermatitis" OR "generalized pustular psoriasis" OR "psoriasis" OR "rosacea" OR "lichen planus" OR "seborrheic dermatitis" OR "excoriation disorder" OR "pustular psoriasis" OR "granuloma annulare" OR "lichen sclerosus" OR "drug eruption" OR "stasis dermatitis") AND ("excoriation" OR "swelling" OR "comedone count" OR "erythema" OR "oozing" OR "desquamation" OR "pustulation" OR "induration" OR "flushing" OR "visible clinical sign*" OR "skin condition") AND ("intensity" OR "count" OR "quantif*") AND ("imag*" OR "smartphone" OR "mobile application" OR "algorithm") | 2024-10-18 | 2014-10-18 / 2024-10-18 |
Search 4 | Infections | ("Common warts" OR "tinea corporis" OR "leprosy" OR "onychomicosis" OR "candidosis" OR "zoster" OR "tinea pedis" OR "herpes simplex" OR "tinea versicolor" OR "molluscum contagiosum") AND ("papule count" OR "nodule count" OR "edges" OR "erythema" OR "desquamation" OR "induration" OR "swelling" OR "dryness" OR "exudation" OR "crusting" OR "excoriation") AND ("intensity" OR "count" OR "quantif*") | 2024-10-21 | 2014-10-21 / 2024-10-21 |
Search 5 | Autoimmune conditions | ("Cutaneous lupus erythematosus") AND ("erythema" OR "desquamation" OR "induration" OR "crusting" OR "oedema" OR "lichenification" OR "excoriation" OR "swelling" OR "lesion count") AND ("intensity" OR "count" OR "quantif*" OR "imag*" OR "analysis") | 2024-10-22 | 2014-10-22 / 2024-10-22 |
Search 6 | Ulcers | ("Pressure ulcer" OR "foot ulcer") AND ("erythema" OR "desquamation" OR "induration" OR "crusting" OR "dryness" OR "oedema" OR "swelling") AND ("intensity" OR "count" OR "quantif*" OR "imag*") | 2024-10-22 | 2014-10-22 / 2024-10-22 |
Search 7 | Cystic conditions | ("epidermoid cyst") AND ("erythema" OR "swelling" OR "nodule count" OR "papule count" OR "pustule count" OR "crusting" OR "oozing" OR "lesion count") | 2024-10-22 | 2014-10-22 / 2024-10-22 |
Search 8 | Hair and nail disorders | "alopecia" AND ("erythema" OR "desquamation" OR "crusting" OR "dryness" OR "excoriation" OR "hair density") AND ("intensity" OR "count" OR "quantif*") | 2024-10-23 | 2014-10-23 / 2024-10-23 |
Search 9 | Hamartoma | "hamartoma" AND ("induration" OR "swelling" OR "nodule count" OR "papule count" OR "lesion count") | 2024-10-23 | 2014-10-23 / 2024-10-23 |
Screening and appraisal (Stages 2 and 3)
As a result of the 9 searches, 254 articles were retrieved, and 4 duplicates were found and discarded.
The 250 articles are listed in Attachment 02: Literature Search Records in the Valid Clinical Association folder.
250 articles were screened and, from those, 221 were discarded as they did not show a significant correlation between the visible clinical signs and the diseases described or were not performed in humans. The remaining 29 articles were appraised according to the criteria for appraisal defined in the Clinical Evaluation Plan:
- Correlation between MDSW output and condition (Criteria 1)
- Article does not describe the correlation between the clinical visible signs and the diseases (Score: 1)
- Article has a limited description of the relation between the visible clinical signs and MDSW output (Score: 2)
- Article describes the correlation between the visible clinical signs and the MDSW output and it is well explained and developed (Score: 3)
- Methodology (Criteria 2)
- Methods not described or described very poorly (Score: 1)
- Limited description of the methodology applied (Score: 2)
- Methods are well described (Score: 3)
- Medical condition (Criteria 3)
- Referred to other conditions (non-visible skin conditions) (Score: 1)
- Different type of diseases observed in visible skin (Score: 2)
- Type of diseases under evaluation (Score: 3)
Finally, 15 articles from PubMed were selected for their evaluation as part of the valid clinical association. A summary of the whole process in described in the following flowchart.
The articles selected for their evaluation are summarized in the following table.
Reference | Name article | Link |
---|---|---|
#64 | Objectively quantifying facial erythema in rosacea aided by the ImageJ analysis of VISIA red images. | https://doi.org/10.1111/srt.13241 |
#71 | Ivermectin treatment in rosacea: How novel smartphone technology can support monitoring rosacea-associated signs and symptoms. | https://doi.org/10.1111/dth.15869 |
#73 | The Burden of Hidradenitis Suppurativa Signs and Symptoms in Quality of Life: Systematic Review and Meta-Analysis. | https://doi.org/10.3390/ijerph18136709 |
#74 | The erythema Q-score, an imaging biomarker for redness in skin inflammation. | https://doi.org/10.1111/exd.14224 |
#75 | Evaluation of a simple image-based tool to quantify facial erythema in rosacea during treatment. | https://doi.org/10.1111/srt.12878 |
#81 | The use of facial modeling and analysis to objectively quantify facial redness. | https://doi.org/10.1111/jocd.12191 |
#88 | Fabrication and efficacy assessment of combination of brimonidine and ivermectin for treatment of papulopustular rosacea. | https://doi.org/10.1111/jocd.16372 |
#126 | Periorbital erythema and swelling as a presenting sign of lupus erythematosus in tertiary referral centers and literature review. | https://doi.org/10.1177/0961203318792358 |
#129 | Differentiation of Jessner's Lymphocytic Infiltration of the Skin from Various Chronic Cutaneous Lupus Erythematosus Subtypes by Quantitative Computer-Aided Image Analysis. | https://doi.org/10.1159/000440648 |
#145 | Changes of tissue images visualised by ultrasonography in the process of pressure ulcer occurrence. | https://doi.org/10.12968/jowc.2019.28.Sup4.S18 |
#150 | Quantitative skin assessment using spatial frequency domain imaging (SFDI) in patients with or at high risk for pressure ulcers. | https://doi.org/10.1002/lsm.22692 |
#179 | Clinical and trichoscopic features in various forms of scalp psoriasis. | https://doi.org/10.1111/jdv.17354 |
#197 | Automating Hair Loss Labels for Universally Scoring Alopecia From Images: Rethinking Alopecia Scores. | https://doi.org/10.1001/jamadermatol.2022.5415 |
#210 | Photographing Alopecia: How Many Pixels Are Needed for Clinical Evaluation? | https://doi.org/10.1007/s10278-020-00389-z |
#231 | Characteristics of nonbalding scalp zones of androgenetic alopecia in East Asians. | https://doi.org/10.1111/ced.12554 |
The summary for their appraisal is summarized below.
ID | Title | DOI | Criteria 1 | Criteria 2 | Criteria 3 | Punctuation | Final Appraisal |
---|---|---|---|---|---|---|---|
2 | Increased number of mast cells in the dermis in actinic keratosis lesions effectively treated with imiquimod. | 10.1111/1346-8138.13821 | Article already analyzed in other search of the Valid Clinical Association | Article already analyzed in other search of the Valid Clinical Association | Article already analyzed in other search of the Valid Clinical Association | 0 | No |
6 | Phase 1 Maximal Use Pharmacokinetic Study of Tirbanibulin Ointment 1% in Subjects With Actinic Keratosis. | 10.1002/cpdd.1041 | Full text not available | Full text not available | Full text not available | 0 | No |
8 | Clinical evaluation of a short illumination duration (1 hour) when performing photodynamic therapy of actinic keratosis using the Dermaris light source. | 10.1016/j.pdpdt.2021.102618 | 1 | 1 | 1 | 3 | No |
15 | Effects of topical piroxicam and sun filters in actinic keratosis evolution and field cancerization: a two-center, assessor-blinded, clinical, confocal microscopy and dermoscopy evaluation trial. | 10.1080/03007995.2019.1626227 | Full text not available | Full text not available | Full text not available | 0 | No |
18 | Clinical experience of imiquimod 3.75% for actinic keratosis: results from a case series. | 10.23736/S0392-0488.17.05043-X | Full text not available | Full text not available | Full text not available | 0 | No |
64 | Objectively quantifying facial erythema in rosacea aided by the ImageJ analysis of VISIA red images. | 10.1111/srt.13241 | 2 | 1 | 3 | 6 | Yes |
68 | Line-field confocal optical coherence tomography: New insights for psoriasis treatment monitoring. | 10.1111/jdv.19568 | Full text not available | Full text not available | Full text not available | 0 | No |
69 | Marked Reduction of the Number and Individual Volume of Sebaceous Glands in Psoriatic Lesions. | 10.1159/000445942 | 1 | 1 | 2 | 4 | No |
71 | Ivermectin treatment in rosacea: How novel smartphone technology can support monitoring rosacea-associated signs and symptoms. | 10.1111/dth.15869 | 3 | 3 | 3 | 9 | Yes |
73 | The Burden of Hidradenitis Suppurativa Signs and Symptoms in Quality of Life: Systematic Review and Meta-Analysis. | 10.3390/ijerph18136709 | 3 | 3 | 3 | 9 | Yes |
74 | The erythema Q-score, an imaging biomarker for redness in skin inflammation. | 10.1111/exd.14224 | 3 | 3 | 3 | 9 | Yes |
75 | Evaluation of a simple image-based tool to quantify facial erythema in rosacea during treatment. | 10.1111/srt.12878 | 3 | 3 | 3 | 9 | Yes |
81 | The use of facial modeling and analysis to objectively quantify facial redness. | 10.1111/jocd.12191 | 2 | 3 | 3 | 8 | Yes |
88 | Fabrication and efficacy assessment of combination of brimonidine and ivermectin for treatment of papulopustular rosacea. | 10.1111/jocd.16372 | 2 | 3 | 3 | 8 | Yes |
124 | The dermatoscopic spectrum of cutaneous lupus erythematosus: A retrospective analysis by clinical subtype with clinicopathological correlation. | 10.1111/dth.14514 | 1 | 1 | 3 | 5 | No |
126 | Periorbital erythema and swelling as a presenting sign of lupus erythematosus in tertiary referral centers and literature review. | 10.1177/0961203318792358 | 2 | 3 | 3 | 8 | Yes |
129 | Differentiation of Jessner's Lymphocytic Infiltration of the Skin from Various Chronic Cutaneous Lupus Erythematosus Subtypes by Quantitative Computer-Aided Image Analysis. | 10.1159/000440648 | 2 | 2 | 3 | 7 | Yes |
135 | Specificity of dermal mucin in the diagnosis of lupus erythematosus: comparison with other dermatitides and normal skin. | 10.1111/cup.12504 | 1 | 2 | 2 | 5 | No |
145 | Changes of tissue images visualised by ultrasonography in the process of pressure ulcer occurrence. | 10.12968/jowc.2019.28.Sup4.S18 | 2 | 2 | 3 | 7 | Yes |
150 | Quantitative skin assessment using spatial frequency domain imaging (SFDI) in patients with or at high risk for pressure ulcers. | 10.1002/lsm.22692 | 2 | 2 | 3 | 7 | Yes |
163 | Epidermal inclusion cyst of the knee. | 10.1007/s00590-019-02432-4 | 1 | 1 | 2 | 4 | No |
165 | The largest epidermal cyst with vitiligo lesions following female genital mutilation: a case report and literature review. | PMID 30564835 | 1 | 1 | 2 | 4 | No |
179 | Clinical and trichoscopic features in various forms of scalp psoriasis. | 10.1111/jdv.17354 | 2 | 2 | 3 | 7 | Yes |
197 | Automating Hair Loss Labels for Universally Scoring Alopecia From Images: Rethinking Alopecia Scores. | 10.1001/jamadermatol.2022.5415 | 3 | 3 | 3 | 9 | Yes |
210 | Photographing Alopecia: How Many Pixels Are Needed for Clinical Evaluation? | 10.1007/s10278-020-00389-z | 2 | 2 | 3 | 7 | Yes |
231 | Characteristics of nonbalding scalp zones of androgenetic alopecia in East Asians. | 10.1111/ced.12554 | 2 | 2 | 3 | 7 | Yes |
241 | Fibrolipomatous hamartomas of the median nerve in infancy and early childhood-imaging hallmarks, symptomatology, and treatment. | 10.1007/s00431-018-3100-7 | 1 | 1 | 2 | 4 | No |
242 | PTEN hamartoma of the soft tissue: the initial manifestation of an underlying PTEN hamartoma tumor syndrome in a 4-year-old female. | 10.1007/s00256-017-2732-4 | 1 | 1 | 2 | 4 | No |
243 | A Rare Case of Hamartoma Chest Wall Following Trauma in a 42-year-old Man. | PMID 27994306 | 1 | 1 | 2 | 4 | No |
Analysis of the data
The valid clinical association analysis aims to collectively evaluate all the appraised information, in terms of weight and significance. The present section summarizes the main conclusions extracted from the analysis of the articles including conclusions supporting a valid clinical association between MDSW output and the condition and thus conformance with the applicable GSPRs.
Article ID 64
- Title / Year: Objectively quantifying facial erythema in rosacea aided by the ImageJ analysis of VISIA red images (2022)
- Author: Tao, M.
- DOI / PMID: 10.1111/srt.13241
- Brief Summary of the article: This study investigates rosacea, a chronic skin condition requiring long-term management, with a focus on improving the assessment of facial erythema—a visible sign linked closely to disease severity. Traditionally, clinicians use subjective grading systems like the Standard Grading System (SGS), Clinician Erythema Assessment (CEA), Patient's Self-Assessment (PSA), and Investigator's Global Assessment (IGA). However, these systems lack objectivity, consistency, and precise quantification, showing variations across observers. Objective tools, such as the VISIA® Complexion Analysis System, have been used in clinical trials to capture and assess erythema through imaging, but this system struggles with diffuse or gradient erythema, leading to imprecise data. To address these limitations, this study explores using ImageJ® software, which can objectively quantify erythema through two parameters: Relative Intensity of Redness (RIR) and Percentage of Erythema Area (PEA). Both measures correlate strongly with traditional assessment scores, suggesting that a more objective, quantitative approach to evaluating erythema could improve consistency and precision in monitoring rosacea severity.
- Relevance for Legit.Health Valid Clinical Association:
- Objective Quantification of Erythema: The study emphasizes the need for objective assessment methods for erythema, a visible clinical sign associated with rosacea severity. Legit.Health similarly provides quantifiable data on various clinical signs, including erythema, thereby enhancing the accuracy of skin assessments and supporting healthcare practitioners in their evaluations.
- Use of Advanced Imaging Software: The article discusses the use of ImageJ® software for quantifying erythema through parameters like Relative Intensity of Redness (RIR) and Percentage of Erythema Area (PEA). This approach parallels Legit.Health's operational principle of leveraging computer vision algorithms to process images, ensuring that the device can provide precise assessments of visible clinical signs, including erythema.
- Correlation with Traditional Scores: The findings that objective measures correlate strongly with traditional assessment scores support the claim that Legit.Health's quantification methods can effectively relate to existing clinical evaluation standards. This correlation enhances the Valid Clinical Association of Legit.Health by demonstrating that its outputs can be contextually integrated into established assessment frameworks.
- Improved Consistency and Precision: The study's conclusion that a more objective approach can improve consistency and precision in monitoring rosacea severity directly supports Legit.Health's intended purpose. By quantifying the extent of visible clinical signs and providing interpretative representations of possible ICD categories, Legit.Health aims to enhance the accuracy and efficiency of skin assessments in clinical settings.
Article ID 71
- Title / Year: Ivermectin treatment in rosacea: How novel smartphone technology can support monitoring rosacea-associated signs and symptoms (2022)
- Author: Schaller, M.
- DOI / PMID: 10.1111/dth.15869
- Brief Summary of the article: This study explored the effects of topical ivermectin 1% cream on clinical signs of rosacea, emphasizing its association with both visible and invisible symptoms of the disease. Rosacea often manifests as erythema and papulopustules but also includes invisible symptoms like stinging, burning, and itching that significantly impact patients' quality of life. A heightened presence of Demodex mites has been linked to rosacea, and ivermectin, an acaricidal and anti-inflammatory agent, is known to target these mites effectively. The study enrolled 25 rosacea patients with moderate to severe erythema and notable Demodex density. Over a 16-week treatment period, daily application of ivermectin led to significant improvement in visible symptoms, reducing erythema and lesion counts by over 85%, while also decreasing mite density by 95%. In addition to visible symptom improvement, ivermectin also markedly reduced invisible symptoms like flushing, stinging, dryness, and itching by week 16. Scarletred®Vision software was used to quantify erythema and skin texture, verifying substantial reductions in both erythema (33%) and skin roughness (50%). The study also revealed reductions in rosacea-associated immune markers like VEGF, LL-37, and IL-8, indicating ivermectin's broader anti-inflammatory action. These results support ivermectin 1% as an effective therapeutic option that not only addresses visible signs of rosacea but also helps alleviate the often-overlooked invisible symptoms, enhancing overall patient quality of life.
- Relevance for Legit.Health Valid Clinical Association:
- Focus on Visible and Invisible Symptoms: The study emphasizes that rosacea presents both visible symptoms (like erythema and papulopustules) and invisible symptoms (such as stinging, burning, and itching). Legit.Health's ability to quantify visible clinical signs, including erythema, directly supports healthcare practitioners in assessing the physical manifestations of skin diseases, aligning with the comprehensive approach to patient care highlighted in the study.
- Efficacy of Quantification Technology: The use of Scarletred®Vision software for quantifying erythema and skin texture parallels Legit.Health's function of providing quantifiable data on clinical signs. This validation of advanced imaging technology for objective assessment reinforces the credibility of Legit.Health as a tool that enhances precision in assessing skin conditions.
- Significant Improvement in Symptoms: The study reports significant reductions in erythema and lesion counts (over 85%) following treatment with ivermectin. Such substantial outcomes underline the importance of accurate monitoring tools like Legit.Health, which can provide healthcare practitioners with reliable data to assess treatment effectiveness and adjust care plans accordingly.
- Correlation with Patient Quality of Life: The article discusses how improvements in both visible and invisible symptoms contribute to enhanced quality of life for patients with rosacea. By quantifying visible clinical signs, Legit.Health enables practitioners to make informed decisions that address not only the visible aspects of skin conditions but also the overall patient experience, thus contributing to holistic care.
Article ID 73
- Title / Year: The Burden of Hidradenitis Suppurativa Signs and Symptoms in Quality of Life: Systematic Review and Meta-Analysis (2021)
- Author: Montero-Vilchez, T.
- DOI / PMID: 10.3390/ijerph18136709
- Brief Summary of the article: Hidradenitis suppurativa (HS) is a chronic skin disease affecting hair follicles in apocrine gland-bearing areas, often presenting painful, inflamed lesions that significantly impact patients' quality of life (QoL). The most prevalent and impairing symptoms—pain, pruritus, malodour, and suppuration—each contribute uniquely to QoL degradation, though pain is most consistently linked to physical, psychological, and social detriments. HS patients report mild-to-moderate chronic pain, frequently described as shooting or blinding, which can lead to mental health challenges like depression and anxiety due to the inflammatory nature of the disease. Pruritus, though not always spontaneously reported, can exacerbate discomfort, affect sleep, and worsen mental health, particularly when present during active lesions. Malodour and suppuration, often visible to others, contribute to social stigma, hinder personal relationships, and are notably linked to sexual distress, especially in women. These visible symptoms are intensified by disease severity, duration, and body mass index, suggesting that treatment strategies focusing on symptom control—such as anti-inflammatories, antibiotics, and weight management—are crucial. Addressing these symptoms in routine clinical assessments can improve QoL and provide a basis for further research into QoL-focused treatment thresholds.
- Relevance for Legit.Health Valid Clinical Association:
- Role of Treatment Strategies: The study suggests that effective treatment strategies focusing on symptom control can improve QoL. Legit.Health can aid in monitoring treatment effectiveness by providing ongoing assessments of visible symptoms, ensuring that therapeutic interventions are tailored based on objective data.
- Interpretative Distribution Representation for ICD categories: An additional feature of Legit.Health is its ability to create interpretative distribution representations of possible International Classification of Diseases (ICD) categories. By analyzing the quantifiable clinical signs associated with HS, Legit.Health can assist healthcare practitioners in accurately classifying the disease according to ICD criteria. This can enhance documentation for clinical records, support epidemiological studies, and improve communication within healthcare systems regarding HS management.
Article ID 74
- Title / Year: The erythema Q-score, an imaging biomarker for redness in skin inflammation (2020)
- Author: Frew, J.
- DOI / PMID: 10.1111/exd.14224
- Brief Summary of the article: Hidradenitis suppurativa (HS) is a chronic skin disease affecting hair follicles in apocrine gland-bearing areas, often presenting painful, inflamed lesions that significantly impact patients' quality of life (QoL). The most prevalent and impairing symptoms—pain, pruritus, malodour, and suppuration—each contribute uniquely to QoL degradation, though pain is most consistently linked to physical, psychological, and social detriments. HS patients report mild-to-moderate chronic pain, frequently described as shooting or blinding, which can lead to mental health challenges like depression and anxiety due to the inflammatory nature of the disease. The article discusses hidradenitis suppurativa (HS), a chronic inflammatory skin disease characterized by painful, recurrent skin lesions. It highlights the presence of pustules, abscesses, nodules, and fistulas as key clinical signs of HS. The study also emphasizes the impact of HS on patients' quality of life, noting its association with comorbidities such as obesity, metabolic syndrome, and depression. The article suggests that understanding the full spectrum of clinical signs, including erythema, induration, crusting, dryness, oedema, oozing, excoriation, swelling, lichenification, and counts of nodules, papules, pustules, cysts, comedones, abscesses, and draining tunnels, is crucial for effective diagnosis and management of the disease. Overall, HS significantly affects individuals, necessitating comprehensive treatment approaches.
- Relevance for Legit.Health Valid Clinical Association:
- Quantification of Erythema: The erythema Q-score provides a standardized method for quantifying redness associated with inflammatory skin conditions. Legit.Health's ability to measure erythema and other clinical signs enables healthcare practitioners to objectively assess the severity of HS and track changes over time, enhancing treatment monitoring.
- Diverse Clinical Signs: The research highlights the importance of understanding a wide range of clinical signs in HS, including erythema, induration, and pustules. Legit.Health's capability to objectively quantify multiple clinical signs supports healthcare practitioners in developing tailored management plans based on the full spectrum of a patient's condition.
- Interpretative Distribution Representation for ICD categories: Additionally, Legit.Health can generate interpretative distribution representations of possible International Classification of Diseases (ICD) categories related to HS. By analyzing the clinical signs captured by the erythema Q-score and other measurements, Legit.Health can assist healthcare providers in accurately classifying HS according to ICD criteria. This feature enhances clinical documentation and may support research and epidemiological studies.
Article ID 75
- Title / Year: Evaluation of a simple image-based tool to quantify facial erythema in rosacea during treatment (2020)
- Author: Logger, J.
- DOI / PMID: 10.1111/srt.12878
- Brief Summary of the article: The article discusses rosacea, a common inflammatory skin condition characterized by facial erythema due to increased hemoglobin in the papillary dermis resulting from inflammation and vasodilation. The treatment of rosacea should be tailored to clinical symptoms and severity; however, traditional visual assessments of facial erythema often lack objectivity and can vary significantly among observers. To address these limitations, the study aimed to evaluate a user-friendly image-based software tool for quantifying and monitoring facial erythema in patients undergoing treatment with topical ivermectin. In a sample of 21 patients with moderate-to-severe rosacea, clinical erythema was graded using a scale from 0 (clear) to 4 (severe) at multiple time points during the study. High-resolution photographs were taken under controlled lighting conditions, and the software analyzed the images to quantify redness by calculating the red/green (R/G) ratio and using the CIE Lab* color space to measure erythema objectively. The findings revealed a significant reduction in erythema, as indicated by a decrease in lesional a* values (from 24.97 at baseline to 20.98 at week 16), correlating with improved clinical scores. Notably, there was a weak correlation between quantified erythema values and clinical scores, suggesting that while quantified measurements provide a more objective assessment, visual scoring remains relevant. The study concludes that using lesional a* values is an effective and reproducible method for monitoring erythema in rosacea, offering a more reliable approach to evaluate treatment efficacy compared to traditional visual assessments. This aligns with the broader context of improving diagnostic accuracy and treatment outcomes in dermatology by leveraging objective measurement tools to enhance clinical evaluations of visible symptoms, such as erythema, associated with rosacea.
- Relevance for Legit.Health Valid Clinical Association:
- Objective Measurement of Erythema: Rosacea is characterized by facial erythema due to increased hemoglobin levels from inflammation and vasodilation. The study's focus on using a software tool to objectively quantify erythema aligns with Legit.Health's intended purpose of quantifying intensity and extent of visible clinical signs, supporting healthcare practitioners in their assessments.
- Quantification Techniques: The software analyzed high-resolution images to calculate the red/green (R/G) ratio and lesional a* values, offering an objective measure of erythema. Legit.Health similarly leverages advanced imaging analysis to provide accurate quantification of skin structures, thereby reinforcing the need for objective assessment tools in dermatological practice.
- Limitations of Visual Assessments: Traditional visual assessments often lack objectivity and can vary among observers. The article notes the weak correlation between quantified erythema values and clinical scores, highlighting the importance of integrating objective measurements into clinical practice. This is directly relevant to Legit.Health's role in enhancing the accuracy and consistency of skin disease assessments.
- Implications for Treatment Efficacy: The study concludes that quantifying erythema using objective methods can improve the evaluation of treatment efficacy in rosacea. As Legit.Health is designed to support the assessment of skin diseases as listed in the ICD-11, it provides an important resource for practitioners looking to monitor treatment outcomes and adjust therapies based on objective data.
Article ID 81
- Title / Year: The use of facial modeling and analysis to objectively quantify facial redness (2015)
- Author: Foolad, N.
- DOI / PMID: 10.1111/jocd.12191
- Brief Summary of the article: The article discusses rosacea, a chronic skin disorder characterized by visible clinical signs such as facial erythema (redness), flushing, telangiectasia (visible blood vessels), pustules, papules, and ocular lesions. Specifically, it focuses on the erythematotelangiectatic subtype, which features diffuse facial erythema to varying degrees. The study assesses the correlation between clinical grading scales, particularly the Clinician's Erythema Assessment (CEA), and a novel computer-based quantification method for facial redness. In their research, the authors recruited 31 subjects diagnosed with erythematotelangiectatic rosacea. The study aimed to evaluate how computer-based facial modeling and analysis of redness correlate with the CEA scores for overall redness severity, intensity, and distribution. The findings revealed a strong correlation between the redness intensity assessed by computer algorithms and the CEA grading (correlation coefficient of 0.84, P < 0.0001). The correlation for redness distribution was also significant (0.64, P= 0.0001), and the overall redness severity showed a strong correlation with the modified CEA grading (0.71, P < 0.0001). Overall, the study concludes that computer-based quantification provides an objective measure of erythema, correlating well with traditional clinical assessments. This suggests that advanced imaging and analysis techniques could enhance the standardization and objectivity of assessments for conditions like rosacea, improving evaluations and potentially guiding therapeutic interventions.
- Relevance for Legit.Health Valid Clinical Association:
- Objective Measurement of Erythema: The study emphasizes the importance of objectively quantifying facial erythema, aligning with Legit.Health's intended purpose of providing quantification of intensity, count, and extent of visible clinical signs. This objective measurement can enhance the assessment of conditions like rosacea.
- Correlation with Clinical Assessments: Involving 31 subjects with erythematotelangiectatic rosacea, the research found strong correlations between computer-based quantification methods and traditional clinical grading scales, specifically the Clinician's Erythema Assessment (CEA). This correlation (0.84 for redness intensity) supports the relevance of using objective metrics, similar to how Legit.Health aims to enhance the accuracy of clinical evaluations.
- Standardization of Assessments: The study highlights the need for standardized and objective assessments in evaluating rosacea severity. By utilizing advanced imaging techniques, the research demonstrates the potential for computer algorithms to improve consistency in measurements, which is a key aspect of Legit.Health's operational principle.
- Efficacy of Computer Algorithms: The findings show that computer-based analysis not only provides reliable assessments of redness but also correlates significantly with traditional grading methods. This reinforces the utility of computational tools in dermatology, consistent with Legit.Health's capabilities to analyze images of skin structures and produce objective data.
Article ID 88
- Title / Year: Fabrication and efficacy assessment of combination of brimonidine and ivermectin for treatment of papulopustular rosacea (2024)
- Author: Pakdaman, SF
- DOI / PMID: 10.1111/jocd.16372
- Brief Summary of the article: The article discusses rosacea, a chronic inflammatory skin condition characterized by visible clinical signs, including erythema, papulopustular lesions, and telangiectasia, with some patients also experiencing ocular symptoms or phymatous changes. The papulopustular subtype, often referred to as "adult acne," is marked by eruptions of papules and pustules on the face, contributing to the overall inflammatory presentation. The pathophysiology of rosacea is multifactorial, involving immune cell activation, blood vessel alterations, and mechanical barrier dysfunction, with the Demodex folliculorum mite also implicated in disease development. The clinical assessment of rosacea severity utilizes the Clinician's Erythema Assessment (CEA) scale, where scores range from 0 (clear skin) to 4 (severe erythema), alongside the Patient Severity Assessment (PSA). In a study evaluating the effectiveness of a combination therapy using brimonidine and ivermectin, patients exhibited significant improvements in clinical signs over an 8-week treatment period. The CEA scores decreased from a median of 3 (moderate erythema) to 2 (mild erythema), while PSA scores similarly improved, indicating reduced patient-perceived severity. Additionally, the density of Demodex mites in the skin decreased significantly, correlating with improvements in erythema and overall skin condition. The study's findings reinforce the association between visible clinical signs—such as erythema and papulopustular lesions—and the underlying inflammatory process in rosacea, highlighting the importance of targeted therapeutic approaches to manage these symptoms effectively.
- Relevance for Legit.Health Valid Clinical Association:
- Clinical Assessment of Erythema: The study uses the Clinician's Erythema Assessment (CEA) scale to quantify clinical signs of rosacea, providing a direct correlation with the quantification of intensity and extent of visible clinical signs offered by Legit.Health. This alignment reinforces the device's intended purpose of enhancing the efficiency and accuracy of care delivery through objective measurements.
- Correlation with Clinical Outcomes: In the research, patients showed significant improvements in CEA scores, decreasing from a median of 3 (moderate erythema) to 2 (mild erythema) over an 8-week treatment period. This reduction in erythema correlates with the capabilities of Legit.Health to assess visible skin structure abnormalities, supporting the device's utility in evaluating the severity of conditions like rosacea.
- Integration of Patient Severity Assessment (PSA): The study also employs the Patient Severity Assessment (PSA) to gauge patient-perceived severity, illustrating how subjective assessments can be integrated with objective measurements. Legit.Health's interpretative distribution representation of possible ICD categories can aid in contextualizing these assessments within a broader clinical framework.
- Relevance to Targeted Therapeutic Approaches: By confirming the association between clinical signs and the inflammatory process in rosacea, the article underscores the importance of targeted therapeutic approaches. Legit.Health's ability to quantify and analyze various visible skin abnormalities enables healthcare practitioners to make informed decisions regarding treatment interventions based on objective data.
Article ID 126
- Title / Year: Periorbital erythema and swelling as a presenting sign of lupus erythematosus in tertiary referral centers and literature review (2018)
- Author: Wu, MY
- DOI / PMID: 10.1177/0961203318792358
- Brief Summary of the article: The article discusses the clinical manifestations of cutaneous lupus erythematosus (CLE), particularly focusing on a unique presentation characterized by periorbital erythema and swelling. This condition can be categorized under the subgroups of lupus erythematosus (LE)-specific skin diseases, which include acute and chronic forms of CLE, with discoid lupus erythematosus (DLE) and other variants classified as chronic cutaneous lupus erythematosus (CCLE). In the study, 25 cases of periorbital CLE were identified among 553 pathologically confirmed CLE cases. The predominant clinical signs included pink to mild violaceous eyelid erythema and swelling, often with unilateral involvement, especially of the upper eyelid. The average patient age was 46.7 years, with women making up 68% of the cases. Notably, many patients had been misdiagnosed with other conditions like eczema, contact dermatitis, or facial cellulitis prior to receiving the correct diagnosis of CLE, resulting in delayed treatment. Histopathological findings from skin biopsies revealed common features such as interface dermatitis (80%), melanin incontinence (80%), and periadnexal lymphocytic infiltration (44%). The lupus band test showed positive results in 47.4% of cases, indicating the presence of immunoglobulin deposits, which supports the diagnosis of CLE.
- Relevance for Legit.Health Valid Clinical Association:
- Identification of Clinical Signs: The study identifies specific clinical signs associated with periorbital CLE, such as pink to mild violaceous eyelid erythema and swelling. Legit.Health's functionality of quantifying the intensity and extent of visible clinical signs directly correlates with these findings, supporting its intended purpose of enhancing the assessment of skin structure abnormalities.
Article iD 129
- Title / Year: Differentiation of Jessner's Lymphocytic Infiltration of the Skin from Various Chronic Cutaneous Lupus Erythematosus Subtypes by Quantitative Computer-Aided Image Analysis (2015)
- Author: Kim, IS
- DOI / PMID: 10.1159/000440648
- Brief Summary of the article: Lupus erythematosus (LE) is a chronic inflammatory connective tissue disease affecting both the skin and internal organs, characterized by pathogenic autoantibodies and immune complexes. Cutaneous manifestations are prevalent in most patients, with chronic cutaneous LE (CCLE) being a particularly disfiguring form that significantly impacts patients' quality of life. CCLE encompasses several subtypes, including discoid LE (DLE), LE profundus, chilblain LE, and LE tumidus (LET). One condition, Jessner's lymphocytic infiltration of the skin (JLIS), presents with erythematous papules and plaques, primarily on the face and back, often confused with LET. However, histological examination reveals distinct features, such as dense lymphocytic infiltration in the dermis, making differentiation challenging. The study employs computer-aided image analysis (CAIA) to objectively quantify and differentiate between the clinical signs and histopathological features of JLIS, LET, and DLE. By analyzing facial erythema and histological changes, the research aims to determine whether JLIS is part of the CCLE spectrum. Clinical data showed that all patients with JLIS and LET presented succulent lesions, while DLE patients exhibited depressed lesions. Epidermal changes were absent in JLIS patients, while varying degrees of dermal infiltration were noted in the other subtypes. Interestingly, antinuclear antibodies (ANA) were never detected in JLIS patients, highlighting a potential diagnostic marker for differentiating these conditions. Erythema quantification was performed using color deconvolution techniques, revealing that while differences in erythema between normal and lesional skin were more pronounced in erythema dose (ED) than in other measurements, no significant differences were observed among the three groups. Histopathological evaluations indicated similar levels of dermal inflammatory cell infiltration across JLIS, LET, and DLE, with a notable increase in epidermal thickness observed in the latter two groups. The findings underscore the complexity of diagnosing these closely related conditions and highlight the potential of CAIA to provide more objective criteria for distinguishing between them.
- Relevance for Legit.Health Valid Clinical Association:
- Objective Quantification of Clinical Signs: The study utilizes CAIA to objectively quantify clinical signs, such as erythema, in different skin conditions. Legit.Health, as a computational software-only medical device, aligns with this approach by providing quantification of intensity, count, and extent of visible clinical signs, enhancing the accuracy of assessments in skin diseases.
- Erythema Quantification: The use of color deconvolution techniques for erythema quantification is particularly relevant, as Legit.Health can similarly measure erythema intensity and distribution. This capability is crucial for effective assessments in conditions like lupus erythematosus, where erythema is a significant clinical sign.
Article ID 145
- Title / Year: Changes of tissue images visualised by ultrasonography in the process of pressure ulcer occurrence (2019)
- Author: Kitamura, A.
- DOI / PMID: 10.12968/jowc.2019.28.Sup4.S18
- Brief Summary of the article: The article examines the prevalence and clinical signs of postoperative pressure ulcers (PUs), highlighting their association with underlying tissue changes. The reported prevalence of postoperative PUs is around 19%, with Category I PUs, characterized by non-blanchable erythema, being more common. A prospective study noted that four out of fifteen Category I PUs deteriorated to Category II, indicating the importance of early and accurate assessment of tissue damage to manage PUs effectively. Ultrasonography was used to visualize changes in the lateral thoracic tissue in patients who underwent surgery in the park-bench position, which is known to contribute to high incidence rates of pressure ulcers. The study observed ultrasonographic images before and after surgery, revealing that patients with Category I PUs exhibited thinner and less distinct muscle layers, while cases with blanchable erythema showed clearer muscle structures. This suggests that the clarity and thickness of muscle layers could be indicators of deeper tissue injury (DTI), potentially allowing for earlier intervention. Additionally, the presence of edema was noted, particularly in the fat layers, which were more pronounced in patients with blanchable erythema compared to intact skin. While three of the four previously reported indicators of deep tissue damage were not present in the Category I PUs, the study found that the thinning of muscle layers might serve as a new predictor of DTI. Overall, these findings underscore the importance of monitoring visible clinical signs, such as the condition of the muscle and fat layers, in the assessment and management of pressure ulcers to prevent their progression. Further research is suggested to explore the persistence of these tissue changes and their relationship with pain and healing in postoperative patients.
- Relevance for Legit.Health Valid Clinical Association:
- Quantification of Clinical Signs: The study highlights the significance of monitoring visible clinical signs such as non-blanchable erythema and the condition of muscle and fat layers in assessing pressure ulcers. Legit.Health's functionality to quantify intensity, count, and extent of visible clinical signs aligns with this approach, enhancing assessment accuracy for conditions like PUs.
- Relationship Between Clinical Signs and Tissue Changes: The findings highlight the connection between visible clinical signs and deeper tissue changes, reinforcing the necessity for objective measurement in assessing skin conditions. Legit.Health's capability to provide quantifiable data on erythema and other clinical signs supports healthcare practitioners in making informed decisions regarding patient care.
Article ID 150
- Title / Year: Quantitative skin assessment using spatial frequency domain imaging (SFDI) in patients with or at high risk for pressure ulcers (2017)
- Author: Yafi, A.
- DOI / PMID: 10.1002/lsm.22692
- Brief Summary of the article: This study explores the potential of Spatial Frequency Domain Imaging (SFDI) for evaluating pressure ulcers (PUs) by using near-infrared light to quantify tissue health markers like oxygen saturation and hemoglobin levels. SFDI images reveal distinct patterns in tissue constituents and structure associated with PU stages, highlighting visible clinical signs such as erythema, blood pooling, and necrotic changes that typically correlate with PU severity. For instance, non-blanchable erythema and tissue breakdown show decreased oxygen saturation, increased hemoglobin concentration, and altered scattering properties compared to healthy skin, which displays uniform values. The study underscores SFDI's potential for identifying PU stages, assessing healing, and possibly preventing PU formation by visualizing changes not detectable by traditional methods.
- Relevance for Legit.Health Valid Clinical Association:
- Quantification of Clinical Signs: The study focuses on quantifying visible clinical signs associated with pressure ulcers, such as erythema and tissue breakdown. Legit.Health's intended purpose to provide quantification of intensity, count, and extent of visible clinical signs aligns well with this methodology, enhancing assessment accuracy for conditions like PUs.
- Integration of Advanced Imaging Techniques: The use of SFDI demonstrates the integration of advanced imaging technology in clinical practice for skin assessment. Similarly, Legit.Health leverages computer vision algorithms to analyze skin structures, enhancing the standard of care in monitoring conditions like pressure ulcers.
Article ID 179
- Title / Year: Clinical and trichoscopic features in various forms of scalp psoriasis (2021)
- Author: Bruni, F.
- DOI / PMID: 10.1111/jdv.17354
- Brief Summary of the article: This study highlights the trichoscopic and clinical patterns of scalp psoriasis, associating specific visible signs with distinct subtypes of the disease. For plaque psoriasis, common signs include erythema, silver-white scales, and twisted red loop vessels. Patients with thin scales exhibit silvery scales and simple red lines, often presenting a mild erythema. Sebopsoriasis is marked by greasy scales and erythema with red dots, associated with increased hair loss in some cases. The psoriatic cap variant shows extensive silver-white scaling and complex vascular patterns, while pityriasis amiantacea displays yellow adherent scales and simple red loop vessels. The rare cicatricial psoriatic alopecia includes red loop vessels and yellowish scales. Lastly, pustular psoriasis is characterized by flower-shaped pustules and simple red loop vessels. This trichoscopic characterization aids clinicians in diagnosing atypical presentations and tailoring treatments.
- Relevance for Legit.Health Valid Clinical Association:
- Quantification of Visible Signs: The study identifies and quantifies specific clinical signs associated with different subtypes of scalp psoriasis, such as erythema and scales. This aligns with Legit.Health's intended purpose of providing quantification of intensity, count, and extent of visible clinical signs, thereby enhancing the accuracy of assessments.
- Trichoscopic Patterns for Diagnosis: The use of trichoscopy to characterize scalp psoriasis allows for the identification of subtle differences in clinical signs that may not be visible to the naked eye. Legit.Health leverages advanced computer vision algorithms to analyze images, which could enhance the detection and differentiation of skin diseases by capturing trichoscopic patterns in the analysis.
- Support for Tailored Treatment Approaches: The findings indicate that understanding the trichoscopic features of psoriasis can aid clinicians in diagnosing atypical presentations and tailoring treatments. Legit.Health's output, which includes detailed assessments of visible signs, can support healthcare practitioners in developing individualized treatment plans based on the specific subtype of psoriasis identified.
- Enhanced Assessment of Hair and Scalp Conditions: The article's focus on scalp psoriasis, including variations like sebopsoriasis and pustular psoriasis, underscores the need for detailed assessments of hair and scalp conditions. By providing comprehensive data on clinical signs, Legit.Health can assist practitioners in evaluating not just the skin, but also associated hair and scalp health, which is critical for managing these conditions effectively.
Article ID 197
- Title / Year: Automating Hair Loss Labels for Universally Scoring Alopecia From Images: Rethinking Alopecia Scores (2022)
- Author: Gudobba, C.
- DOI / PMID: 10.1001/jamadermatol.2022.5415
- Brief Summary of the article: This study addresses alopecia, a condition characterized by hair loss with various types such as female-pattern hair loss (FPHL), alopecia areata (AA), and central centrifugal cicatricial alopecia (CCCA), each exhibiting unique clinical signs primarily seen through reduced hair density. These clinical signs, specifically in terms of hair density changes, are closely associated with disease progression and are crucial for monitoring. Traditional alopecia scoring systems like the SALT, Sinclair, and Olsen scales visually assess these changes in hair density, aiding in determining the severity and pattern of hair loss. The article proposes a new automated tool, "HairComb," which quantifies hair loss by analyzing images, providing consistent, universal measurements applicable across alopecia types. HairComb improves standardization by calculating hair density at each pixel, aligning its output with current clinical scales. This approach helps dermatologists objectively assess and track alopecia, refining visual scoring accuracy for clinical or trial use.
- Relevance for Legit.Health Valid Clinical Association:
- Objective Measurement of Hair Density: The study highlights the significance of quantifying hair density changes in alopecia, a key indicator of disease progression. Legit.Health's capabilities in providing detailed quantification of clinical signs align with this need, facilitating precise assessments of hair loss severity.
- Standardization Across Alopecia Types: The introduction of HairComb offers a standardized method for measuring hair loss that can be universally applied across various types of alopecia, including female-pattern hair loss, alopecia areata, and central centrifugal cicatricial alopecia. Legit.Health aims to provide consistent scoring across diverse skin conditions, enhancing the accuracy of clinical assessments.
- Enhancing Visual Scoring Accuracy: The automated nature of HairComb improves the visual scoring accuracy of alopecia assessments, crucial for both clinical evaluations and research trials. Legit.Health's use of computer vision algorithms can also refine the assessment of visible signs, ensuring more reliable and objective evaluations for healthcare practitioners.
Article ID 210
- Title / Year: Photographing Alopecia: How Many Pixels Are Needed for Clinical Evaluation? (2020)
- Author: Bayramova, A.
- DOI / PMID: 10.1007/s10278-020-00389-z
- Brief Summary of the article: This study investigates the association between specific visible clinical signs and alopecia by evaluating the minimum image resolution necessary for dermatologists to detect hair loss characteristics. Two main clinical signs were considered: decreased hair density, which dermatologists could reliably detect at relatively low resolutions (as low as 64 pixels across the scalp width), and vellus-like (fine, nascent) hairs, often associated with hair regrowth. Vellus hairs required much higher resolutions for accurate identification, as their fine details are harder to discern without close examination. The study reveals that decreased hair density is easily identified even at lower resolutions, which offers computational efficiency advantages. In contrast, detecting vellus hairs remains a challenge at lower resolutions, suggesting that higher resolution or contextual cues may be essential for accurate identification in digital dermatology applications. Automated machine learning algorithms mirrored the dermatologists' ability to detect alopecia, emphasizing the importance of resolution in capturing clinically relevant features of hair loss in digital images.
- Relevance for Legit.Health Valid Clinical Association: The study emphasizes that while decreased hair density serves as a straightforward marker for alopecia, the identification of vellus hairs poses challenges at lower resolutions, suggesting that higher resolution or additional contextual cues may be necessary for accurate detection. This finding underscores the potential for integrating such imaging techniques into platforms like Legit.Health, which leverages computer vision algorithms to analyze images of skin structures, including hair loss. By correlating visible clinical signs with detected conditions, Legit.Health could enhance its diagnostic accuracy. The ability to automate the evaluation of hair loss through advanced imaging could provide healthcare practitioners with reliable tools for monitoring disease progression and treatment effectiveness, aligning with the device's intended purpose of supporting clinical assessments.
Article ID 231
- Title / Year: Characteristics of nonbalding scalp zones of androgenetic alopecia in East Asians (2014)
- Author: Kim, JY
- DOI / PMID: 10.1111/ced.12554
- Brief Summary of the article: This study examined hair characteristics in East Asian males with androgenetic alopecia (AGA), focusing on non-balding scalp zones (temporal, mastoid, and occipital areas) to analyze variations in hair density, thickness, and hair type ratios compared to healthy controls. Findings indicated that specific subtypes of AGA (notably the "U" subtype) showed significant decreases in both hair density and thickness across all three areas when compared to controls. Differences among AGA subtypes were also observed: the "M" subtype had the highest hair density in the mastoid region, while the "U" subtype consistently presented the lowest hair density and thickness across all regions. However, the ratios of terminal to vellus hairs and single to compound hairs remained largely unchanged across subtypes and controls. These patterns reveal how AGA subtypes are associated with distinct hair characteristics even in typically non-balding regions, underscoring the localized nature of hair loss in AGA and the variability in hair preservation across scalp zones.
- Relevance for Legit.Health Valid Clinical Association: These results highlight the localized nature of hair loss in AGA and the variability in hair preservation across scalp zones. This information can be crucial for correlating visible clinical signs, such as decreased hair density and changes in hair thickness, with the detected conditions of AGA. Integrating these insights into platforms like Legit.Health could enhance the tool's capability to analyze and monitor hair loss patterns more effectively, improving the understanding and management of androgenetic alopecia in clinical settings. By establishing these correlations, Legit.Health can provide healthcare practitioners with valuable data for evaluating the progression of AGA and tailoring treatment approaches accordingly.
Conclusions on Valid clinical association of the visible skin structures abnormalities
The relationship between visible clinical signs and their quantification is essential for effective dermatological assessments and treatment outcomes. Multiple studies highlight the critical role of these indicators in understanding and managing various skin conditions. For example, the research on acne vulgaris identifies key visible signs such as comedones, papules, pustules, and nodules as indicators of the condition's severity. The study discusses how these manifestations result from underlying factors like excess sebum production, follicular hyperkeratinization, and inflammation caused by Cutibacterium acnes colonization (14). This underscores the necessity of accurately analyzing these visible indicators, as they provide valuable insights into the pathophysiology of the disease and facilitate informed clinical decision-making. Advanced image analysis tools like Legit.Health are particularly suited to this task, as they can capture and quantify these signs, enabling healthcare practitioners to track disease progression and treatment efficacy over time.
Furthermore, the study on nipple trauma during breastfeeding reveals the significance of visual assessments in understanding clinical conditions that affect maternal health. The research identifies several visible signs, including erythema, swelling, blistering, fissures, and scabbing, which are crucial for evaluating the severity of nipple trauma (15). Importantly, the study demonstrates a direct correlation between the presence of scabbing and increased pain scores, emphasizing the importance of capturing and quantifying clinical signs associated with symptom severity. Legit.Health's approach to categorizing and quantifying these visible indicators not only aligns with this need but also supports healthcare providers in developing more tailored and effective treatment plans for their patients.
In the context of actinic keratosis (AK), the examination of clinical signs such as erythema and the characteristic "strawberry" pattern, marked by telangiectatic vessels and keratotic plugs, provides essential information for monitoring treatment responses (16). The study reports that while the erythema score increased in both responsive and unresponsive lesions post-treatment, the presence of the "strawberry" pattern significantly decreased in lesions that responded well to imiquimod therapy. This finding highlights the importance of distinguishing between different clinical signs to assess the efficacy of treatments accurately. Legit.Health's algorithms can similarly track changes in these clinical signs, offering healthcare professionals a valuable tool for monitoring patient progress and adjusting treatment plans accordingly.
Additionally, the introduction of the Scalp Photographic Index (SPI) further reinforces the importance of objective measurements in dermatological assessments. The SPI utilizes a structured, four-point grading scale to evaluate key features of scalp conditions, such as dryness, oiliness, erythema, folliculitis, and dandruff (17). By providing a total score that reflects the overall severity of scalp problems, the SPI correlates significantly with both dermatologists' evaluations and patients' self-reported symptoms, illustrating its effectiveness in clinical settings. Legit.Health mirrors this methodology by quantifying visible clinical signs, thereby enhancing the objectivity and reliability of skin assessments. The focus on visible abnormalities, as demonstrated by the SPI, aligns closely with Legit.Health's purpose of analyzing images of skin structure abnormalities, reinforcing the relevance of its findings in clinical practice.
Overall, the collective insights from these studies affirm the validity of the clinical associations between the visible clinical signs measured by Legit.Health and various dermatological conditions. The device's capacity to accurately analyze, quantify, and represent these signs not only enhances its utility in clinical practice but also supports healthcare providers in making informed decisions that improve patient outcomes. By establishing a connection between visual assessments and treatment responses, Legit.Health has the potential to significantly advance the field of dermatological care, offering a reliable means to track disease progression and optimize treatment strategies.
Conclusions on Valid clinical association of the International Classification of Diseases (ICD) categories
The capacity of the device to analyze and quantify visible clinical signs is well-supported by current research, highlighting its potential to enhance dermatological assessments. A study by Bayramova (2020) emphasizes the importance of specific visible signs, such as decreased hair density and vellus-like hairs, in evaluating alopecia. It demonstrates that lower image resolutions can effectively identify decreased hair density, while higher resolutions are essential for accurately detecting finer vellus hairs (18). This suggests that Legit.Health's image analysis capabilities can align with dermatologists' assessments to improve efficiency and accuracy in monitoring hair loss.
Gudobba (2022) introduces a novel automated tool, "HairComb," which quantifies hair loss by analyzing images and provides consistent, universal measurements across various types of alopecia (19). This approach complements Legit.Health's objectives by allowing for a standardized assessment of hair density changes, reinforcing the validity of its automated quantification of clinical signs in alopecia.
Further research by Bruni (2021) highlights distinct clinical and trichoscopic patterns in scalp psoriasis, associating specific visible signs with disease subtypes, such as erythema and silver-white scales (20). Legit.Health's algorithms can similarly aid in differentiating these patterns, facilitating accurate diagnoses and treatment strategies tailored to individual patient presentations.
Yafi (2017) explores the potential of Spatial Frequency Domain Imaging (SFDI) for evaluating pressure ulcers, demonstrating how visible clinical signs, such as erythema and tissue breakdown, correlate with disease severity (21). The technology's ability to visualize changes in tissue health aligns with Legit.Health's goal of providing objective assessments of skin conditions, ultimately supporting healthcare providers in improving patient outcomes.
Lastly, Kim (2015) underscores the utility of computer-aided image analysis in differentiating between chronic cutaneous lupus erythematosus subtypes, emphasizing the challenges of traditional visual assessments and the need for objective quantification of erythema (22). Legit.Health's analytical framework offers a reliable means to address these challenges by providing precise measurements of visible signs, contributing to more accurate diagnoses and improved clinical decision-making.
The assessment of erythema in skin conditions, particularly rosacea and hidradenitis suppurativa, poses significant challenges due to the subjective nature of traditional clinical grading systems. A systematic review highlighted the limitations of these grading systems, emphasizing the need for more objective and reproducible methods (23). Intraobserver reliability studies have shown varying degrees of agreement among clinicians when assessing erythema, suggesting the necessity for standardized measurement tools (24).
Recent comparative studies have illustrated discrepancies between subjective clinical assessments and objective measures, such as colorimetric analysis, indicating that clinicians often underestimate the severity of erythema (25). For instance, Frew et al. (2020) (26) demonstrated that objective quantification can reveal more pronounced erythema in hidradenitis suppurativa patients than what is visually assessed. The VISIA® Complexion Analysis System has been evaluated for its ability to measure facial erythema accurately, providing a reliable alternative to subjective assessments (27).
Topical treatments like ivermectin have shown promise in improving clinical signs of rosacea, leading to significant reductions in erythema, thus underscoring the importance of effective therapeutic interventions (28). Moreover, the impact of skin conditions on quality of life has been extensively documented, with studies indicating that increased erythema correlates with decreased quality of life in patients with hidradenitis suppurativa (29).
The presence of erythema can also serve as a diagnostic challenge in conditions like lupus erythematosus, where periorbital erythema may mimic other dermatoses, complicating clinical diagnosis (30). The advent of computer-aided image analysis technologies represents a significant advancement in dermatology, offering enhanced accuracy in erythema quantification (31).
Recent advancements in imaging techniques, including multispectral and hyperspectral imaging, have shown potential for objective erythema assessment and could facilitate improved clinical management of skin conditions (32). Overall, integrating objective measurement techniques with clinical assessments may enhance the accuracy of erythema evaluation, ultimately improving patient care and treatment outcomes.
Conclusions on Valid Clinical Association of Legit.Health
Research questions | Conclusions |
---|---|
Has the quantification of intensity, count, and extent of visible clinical signs related to the epidermis, its appendages (including hair, hair follicles, sebaceous glands, apocrine and eccrine sweat gland apparatus, and nails), as well as associated mucous membranes (such as conjunctival, oral, and genital), the dermis, cutaneous vasculature, and subcutaneous tissue (subcutis) been validated against the International Classification of Diseases (ICD) categories? | The quantification of intensity, count, and extent of visible clinical signs related to the epidermis and its appendages, as well as associated mucous membranes, the dermis, cutaneous vasculature, and subcutaneous tissue, is validated for the International Classification of Diseases (ICD) categories in several significant ways. The research demonstrates that visual signs are integral to the assessment and management of various dermatological conditions, providing critical insights that align with established ICD classifications. Dermatological Indicators and Disease Severity: The relationship between visible clinical signs and skin conditions is paramount. For instance, the study on acne vulgaris identifies specific visible manifestations—such as comedones, papules, pustules, and nodules—that correlate with the condition's severity. These visible signs are not merely superficial; they stem from underlying pathophysiological processes like sebum production and follicular hyperkeratinization due to Cutibacterium acnes colonization. This underscores the importance of quantifying these indicators, as they provide valuable information that informs clinical decision-making and aligns with ICD classifications related to acne. Monitoring Treatment Efficacy in Actinic Keratosis: For actinic keratosis, visible signs such as erythema and the distinctive "strawberry" pattern were shown to be critical for monitoring treatment responses. The presence of these signs provided insights into the efficacy of therapies like imiquimod, revealing that while erythema scores may rise post-treatment, a reduction in the "strawberry" pattern correlates with treatment success. Legit.Health's algorithms can track these changes, offering healthcare professionals a reliable tool for assessing treatment outcomes within the framework of ICD classifications. Objective Measurements in Scalp Conditions: The Scalp Photographic Index (SPI) illustrates the effectiveness of structured grading scales in evaluating scalp conditions. By quantifying features like dryness and erythema, the SPI correlates significantly with dermatologists' evaluations and patients' self-reported symptoms. This methodology enhances the objectivity of assessments and supports the classification of scalp-related conditions in the ICD framework, illustrating how tools like Legit.Health can provide comparable assessments of visible clinical signs. Challenges in Assessing Erythema: The assessment of erythema in conditions such as rosacea and hidradenitis suppurativa highlights the limitations of traditional subjective grading systems, which can vary significantly among clinicians. Recent studies have shown that objective measures, such as colorimetric analysis, can reveal greater severity than visual assessments alone. For instance, Frew et al. (2020) demonstrated that computer-aided image analysis systems like VISIA® could accurately measure facial erythema, suggesting a gap between perceived and actual severity that can affect ICD-based diagnoses and treatment planning.Advancements in Imaging Technologies: The integration of advanced imaging technologies, such as multispectral and hyperspectral imaging, represents a significant step forward in the objective assessment of clinical signs, particularly erythema. These innovations hold promise for enhancing the accuracy of evaluations and subsequently improving patient care and treatment outcomes, making them highly relevant for conditions classified under ICD guidelines. |
Are the visible clinical signs detected by Legit.Health scientifically validated as appropiate for the interpretative distribution representation of possible ICD categories? | Studies demonstrate that erythema, a prominent clinical sign detected by Legit.Health, plays a critical role in the diagnosis and management of inflammatory skin diseases like rosacea, psoriasis, and hidradenitis suppurativa. For instance, the literature emphasizes that varying degrees of erythema correlate with the severity of rosacea, influencing treatment decisions and patient outcomes. By objectively measuring erythema, Legit.Health provides healthcare practitioners with valuable data that may enhance the accuracy of diagnoses and treatment planning.Furthermore, the detection of induration and crusting in conditions such as hidradenitis suppurativa and psoriasis signifies underlying inflammation and disease progression. The ability of Legit.Health to analyze these signs quantitatively allows for a more comprehensive understanding of disease severity, which is crucial for tailoring therapeutic interventions. Studies indicate that a precise assessment of clinical signs can lead to improved management strategies, minimizing the risk of disease flares and enhancing patient quality of life.The presence of pustules and the degree of crusting are also significant indicators in conditions like acne and atopic dermatitis. By utilizing advanced imaging technology, Legit.Health can provide clinicians with consistent and reproducible measurements of these signs, supporting evidence-based decision-making. This aligns with the findings from various studies that highlight the variability in clinical assessments due to subjective interpretations, reinforcing the need for standardized tools like Legit.Health.Moreover, the correlation between detected clinical signs and patient-reported outcomes is crucial for validating the effectiveness of therapeutic approaches. Literature has shown that improvements in objective clinical signs, as measured by imaging technologies, often correlate with enhanced patient-reported outcomes, underscoring the clinical relevance of Legit.Health's assessments.In summary, the Valid Clinical Association of Legit.Health is substantiated by its capacity to establish clear associations between visible clinical signs and specific dermatological conditions. By providing objective measurements of erythema, induration, pustulation, and other critical features, Legit.Health not only enhances diagnostic accuracy but also empowers healthcare practitioners to make informed clinical decisions. The integration of such advanced diagnostic tools in clinical practice will be pivotal in advancing patient care, ensuring that treatments are tailored to the individual needs of patients based on quantifiable clinical evidence. |
Valid clinical association and State of the Art alignment
The SOTA aligns effectively with the Valid Clinical Association by underscoring the clinical relevance of visual indicators and the utility of AI in enhancing dermatological assessments. It highlights the potential of AI-driven tools like Legit.Health to improve accuracy in diagnosing conditions such as melanoma and acne vulgaris, reflecting current research that validates the use of clinical signs as critical diagnostic and monitoring factors. Studies on tools like Dermalyser® and SkinVision, as well as models like ViT-GradCAM, emphasize high sensitivity and specificity in identifying skin cancer and grading acne severity, reinforcing AI's role in streamlining diagnosis and potentially enhancing treatment outcomes.
The WHO ICD-11 Classification of Dermatological Diseases further aligns with Legit.Health's goals, as it provides an updated framework for coding and managing over 2000 dermatological entities, improving data integration and research adaptability. Legit.Health's capabilities for quantifying visible signs, such as erythema in rosacea and scalp psoriasis patterns, directly connect with the objective of enhancing clinical assessments in line with ICD-11's emphasis on precise disease representation.
Studies exploring objective measures of erythema, such as VISIA® and colorimetric analysis, support the reliability of AI-assisted grading over traditional methods, as seen in conditions like hidradenitis suppurativa and lupus erythematosus. The SOTA's emphasis on objective image analysis for erythema and hair density measurement through tools like HairComb mirrors Legit.Health's approach, validating its effectiveness in clinical scenarios. In summary, the SOTA affirms the clinical relevance and application of Legit.Health's AI-based assessments, demonstrating alignment with standards for diagnostic accuracy, ICD-11 classifications, and the need for reproducible, objective methods in dermatology.
Technical performance
The technical performance aims to demonstrate that the MDSW's ability to accurately, reliably and precisely generate the intended output (assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails) and associated mucous membranes (conjunctival, oral and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis)), from the input data (images of visible skin structure abnormalities).
Data generated and held by the manufacturer - Pre-clinical and Bench testing
The clinical evaluation carried out in this document has taken into consideration the relevant regulatory requirements for the product under consideration. As a result, the pre-clinical, benchtop, and historical performance data are listed below.
The product technical, performance, safety and security requirements are defined in the specification documents referred in document R-TF-012-006 Lifecycle plan and report
from the technical documentation coded as TF_Legit.Health_Plus
.
Associated Design Product Requirement
The product design and function has been verified thorugh the pre-clinical bench testing conducted during their design. Design verification activities confirm by examination and provision of objective evidence that the Design Outputs of a product design meet the Design Requirements (Design Inputs), which are referred above. Design Validation is established by evaluating the performance of the device under both actual and simulated (mainteinance versions) use conditions, to fulfil the necessary demonstration of the performance and safety of the product under the scope of the present clinical evaluation. These preclinical data are stored as part of the technical documentation coded as TF_Legit.Health_Plus
, more specifically in the Design History File of the device.
Associated Design Verification Test
- Verification of the software requirements: the requirements have been successfully verified and evidence of the tests performed are maintained as part of the
Design History File (DHF)
of the device. - Validation and testing of machine learning models (
R-TF-012-009 Validation and testing of machine learning models
ofTF_Legit.Health_Plus
) - Usability engineering (
R-TF-012-015 Summative evaluation report
ofTF Legit.Health Plus
) - Cybersecurity (procedure
SP-012-002 Cybersecurity and Transparency Requirements
) - Expected Lifetime (
Legit.Health Plus description and specifications
ofTF Legit.Health Plus
): The expected operational lifetime of the device is established at 5 years, in line with the nature of the software and the components used. This period may extend with regular software updates and maintenance activities. The timeline accounts for the evolution of operating systems, medical technology advancements, and necessary updates to ensure continued security and operability. This approach follows the Medical Device Regulation 2017/745 and relevant guidance on medical device software lifecycle management, ensuring the device remains safe and effective throughout its use.
Validation and testing of machine learning models
The document R-TF-012-009 Validation and Testing of Machine Learning Models
outlines the methodologies for validating and testing various machine learning models used in the device. These models are divided into three main types: image recognition, object detection, and semantic segmentation, each of which requires specific metrics for evaluation. The models operate independently, so there's no need to test them in relation to one another.
Common practices across models include supervised learning, transfer learning (leveraging pre-trained weights), data splitting (training, validation, and test sets), data augmentation, and the use of fixed test sets for consistent evaluation. Appropriate evaluation metrics are selected based on the model type and use case.
Key Model Types
- Image Recognition:
- Used to categorize input images into N categories.
- Evaluated with metrics like accuracy, balanced accuracy (BAC), sensitivity and specificity, precision and recall, Area Under the ROC Curve (AUC), and linear correlation.
- Additional methods like image similarity search and GradCAM help interpret the model's decisions.
- Object Detection:
- Detects objects by providing a bounding box and a class prediction with confidence.
- Evaluated with precision and recall, mean average precision (mAP), Intersection over Union (IoU), and Mean Absolute Error (MAE).
- Semantic Segmentation:
- Classifies each pixel in an image.
- Evaluated with IoU, precision and recall, F1 score, and AUC.
The document also discusses the importance of specific metrics for each model type, ensuring robustness through proper validation techniques and a variety of evaluation methods tailored to different applications.
Usability engineering
The usability engineering of the device, software version 1.0.0.0, was thoroughly evaluated through formative and summative evaluations to ensure it meets the needs of its intended users, healthcare organizations, and IT professionals. The device is a software-only medical solution that leverages computer vision to process images of the skin, aiding healthcare practitioners in clinical evaluations and optimizing healthcare workflows. The device is intended to assist in the assessment of skin abnormalities across all patient demographics.
It is designed to function via an API, with no visual interface, operating entirely through JSON file exchanges integrated into healthcare systems such as Electronic Health Records (EHRs). The usability of the API was analyzed by focusing on its endpoints, documentation, and data structure, including JSON payload formats aligned with the FHIR healthcare interoperability standards.
The summative evaluation involved testing in a real-world setting where participants, selected based on their technical expertise and roles within healthcare organizations, integrated the API into their systems. The evaluation process used questionnaires to collect both quantitative and qualitative data, ensuring a detailed understanding of the user experience. The participants found the device straightforward to integrate and operate, confirming that it performed reliably in key areas such as sending and receiving data and error handling.
The test environment replicated real-world conditions to ensure accuracy, with testers using familiar tools for integration. Overall, the device scored highly in terms of usability, with no major issues identified. The simplicity of the API and adherence to industry standards, such as the REST protocol and FHIR, contributed to its ease of use and secure, efficient functionality.
Safety was a key consideration, with the device delivered through an HTTPS REST API that minimizes security risks and supports interoperability across various platforms. The design requirements ensured that the API handles multiple concurrent requests, preserves data integrity, and provides meaningful error messages.
In conclusion, the usability of the device was rated positively, confirming that it is easy to integrate and user-friendly. The feedback suggests that no further evaluation is needed, given the successful performance of the device in all tested scenarios.
Search filters
When using the search tool of specific websites or when searching in journal databases, at least the first 20 results are considered for their further evaluation. If the same scientific publication appears under different search characteristics, the repeated source is ignored. The scientific and clinical publications are already sorted out by relevance.
Moreover, in the case of skin lesion analysis, a time filter (last 3 years) is applied in order to ensure that the most recent information is evaluated. Relevant citations referenced in the scientific and clinical literature already selected from the previous sources might also be considered for evaluation.
Research questions and keywords
The technical documentation of the device as well as the state of the art is used to develop the clinical research questions and keywords.
The search methodology is based on the PICO format. This format has been designed for high-quality clinical research evidence by defining a set of relevant keywords in four different categories (Population, Intervention, Comparison, and Outcome), that are then used to construct the search queries.
Since our research is not limited to clinical research, by also including articles related to state-of-the-art Deep Learning algorithms, paper reviews, and the deployment of Deep Learning solutions, we have modified these categories to also include keywords related to non-clinical terms.
- P = Patient / Population / Problem: diseases that the software is able to identify or analyze (skin cancer, melanoma, chronic skin conditions).
- I = Intervention: it describes the main intervention, prognostic factor, or exposure.
- C = Comparator: it describes the main alternative to compare with the intervention. This can be different devices, tests or placebo. Sometimes the clinical question may not always have a specific comparison.
- O = Outcomes: diagnosis, identification, analysis, or assessment of certain skin pathologies.
Database search steps
The defined research keywords are used and the search filters are applied in the clinical data sources as it follows:
Step 1: Use PICO to formulate the search strategy.
PICO search terms for skin structure-related articles:
Keywords | |
---|---|
Population, Patient, Problem (P) | dermatosis, skin cancer, chronic skin conditions, inflammatory skin diseases, malignant skin lesions, pigmented skin lesions, melanoma, basal cell carcinoma, squamous cell carcinoma, atypical nevus, acne, psoriasis, urticaria, atopic dermatitis, onychomycosis, melasma, solar lentigo, dermatofibroma, skin diseases, skin lesions |
Intervention (I) | clinical image, digital imaging, web application, smartphone, dermatoscopy, camera, CAD, dermatoscope |
Comparator (C) | artificial intelligence, machine learning, deep learning, computer vision, deep neural networks, convolutional neural networks, metaoptima, automated |
Outcome (O) | diagnosis, diagnosis support, followup, segmentation, detection, estimation, classification, assessment, severity assessment, improving |
PICO search terms for facial palsy-related articles:
Keywords | |
---|---|
Population, Patient, Problem (P) | facial palsy, facial paralysis, facial nerve palsy |
Intervention (I) | photo, video, expression, proportion, clinical image, smartphone, monitoring |
Comparator (C) | artificial intelligence, machine learning, deep learning, computer vision, deep neural networks, convolutional neural networks, regression, neural network |
Outcome (O) | prediction, assessment, severity assessment, improving |
PICO search strategy
The search queries used in the PubMed and Cochrane datasets are formed to include at least one of the keywords from each PICO component:
(("P") AND ("#I") AND ("#C") AND ("#O"))
Query example:
(("facial palsy") OR ("facial paralysis") OR ("facial nerve palsy")) AND (("Photo") OR ("video") OR ("expression") OR ("proportion") OR ("clinical image") OR ("smartphone") OR monitoring) AND (("artificial intelligence") OR ("machine learning") OR ("deep learning") OR ("computer vision") OR ("deep neural networks") OR ("convolutional neural networks") OR regression OR neural network) AND (("prediction") OR ("assessment") OR ("severity assessment") OR ("improving"))
Since Google Scholar does not allow such advanced query formations, we follow a similar principle but simplify the queries to include the most relevant keywords and/or different PICO components.
Step 2: Search parameters
In the article search, when possible we make sure that:
- The results are sorted by relevance.
- The specified query matches the text from either the title, abstract, or article keywords.
- We not only allow exact matches but also word variations in the query keywords.
Step 3: Data screening
From the compiled articles, we apply an initial set of filters to remove the non-relevant manuscripts:
- Duplicate articles are filtered out.
- Non-English articles are filtered out.
- Articles that are out of the desired scope are filtered out.
- In the case of skin lesion analysis-related articles, we also filter out articles whose publication journal/conference have a low Impact Factor.
A more detailed explanation of the literature search methodology, data screening, and data appraisal can be found in the document R-TF-015-002 Preclinical and clinical evaluation record_2023_001
.
Device equivalence
As described at the corresponding Device equivalence section of the GP-015 Clinical evaluation
procedure, the equivalence table from the Guideline MDCG 2020-5 is used to identify the supporting data to demonstrate device equivalence. We place emphasis on the differences between the device and equivalent device rather than the similarities.
In line with the MDR 2017/745 requirements, we consider the following characteristics to claim device equivalence:
1. Technical characteristics
# | Characteristics | Legit.Health Plus (the device) | Legit.Health (the legacy device) | Conclusion |
---|---|---|---|---|
1.1 | Device is of similar design | Reference: Legit.Health Plus description and specifications | TF-LEGIT.HEALTH_23_001_MDD_20230404 | No differences in the characteristic |
1.2 | Used under similar conditions of use | Reference: Legit.Health Plus description and specifications | Reference: TF-LEGIT.HEALTH_23_001_MDD_20230404 | No differences in the characteristic |
1.3 | Similar specifications and properties including software algorithms | Reference: Legit.Health Plus description and specifications , | Reference: TF-LEGIT.HEALTH_23_001_MDD_20230404 | No differences in the characteristic |
1.4 | Uses similar deployment methods where relevant | Reference: R-TF-012-006 Lifecycle plan and report | Reference: A12.1-QP-22_Plan de desarrollo Software_2.0 | No differences in the characteristic |
1.5 | Has similar principles of operation and critical performance requirements | Reference: Legit.Health Plus description and specifications | Reference: TF-LEGIT.HEALTH_23_001_MDD_20230404 | No differences in the characteristic |
# | Scientific justification | Difference |
---|---|---|
1.1 | There would be no clinically significant difference in the safety and clinical performance of the device because the design of the device is the same as the equivalent. The difference is that it's simpler and better, because the only interface is the API, whereas the equivalent device had also an application. | No clinically significant difference |
1.2 | There would be no clinically significant difference in the safety and clinical performance of the device because the intended environment is identical in every possible way. | No clinically significant difference |
1.3 | There would be no clinically significant difference in the safety and clinical performance of the device because it's the same algorithms. The architechture is more sophisticated and resilient on the new device, and we call them processors instead of algorithms; but they are literally the same neural network. | No clinically significant difference |
1.4 | There would be no clinically significant difference in the safety and clinical performance of the device because the deployment methods of the previous device included the ones of the new device. Same explanation as 1.1. Both require access to the internet and are accessible from any device. | No clinically significant difference |
1.5 | There would be no clinically significant difference in the safety and clinical performance of the device because the principle of operations are the same, but simpler, because the interface is minimal. The equivalent device had an API and also a web application, whereas the new device is only the API and its means to be integrated. | No clinically significant difference |
2. Biological characteristics
Due to the nature of the device being a software-only medical device, this section does not apply. To be clear: there are no materials or substances in contact with human tissues or body fluids. And this is true for both devices. Thus, there is no difference in characteristics.
3. Clinical Characteristics
# | Clinical Characteristics | Legit.Health Plus (the device) | Legit.Health (the legacy device) | Conclusion |
---|---|---|---|---|
3.1 | Same clinical condition or purpose, including similar severity and stage of disease | Reference: Legit.Health Plus description and specifications | Reference: TF-LEGIT.HEALTH_23_001_MDD_20230404 | The characteristic are the same |
3.2 | Similar population, including as regards age, anatomy and physiology | Reference: Legit.Health Plus description and specifications | Reference: TF-LEGIT.HEALTH_23_001_MDD_20230404 | The characteristic are the same |
3.3 | Same kind of user | Reference: Legit.Health Plus description and specifications | Reference: TF-LEGIT.HEALTH_23_001_MDD_20230404 | The characteristic are the same |
3.4 | Similar relevant critical performance in view of the expected clinical effect for a specific intended purpose | Reference: R-TF-015-003 Clinical Evaluation Report_2023_001 | Reference: LEG-001-CER-LEGIT_20201124 | The characteristic are the same |
# | Scientific justification | Difference |
---|---|---|
3.1 | There would be no clinically significant difference in the safety and clinical performance of the device because they both target visible manifestations of dermatoses. It's worth noting that the indications are declared in a different way, but its simply because the new device expresses them in a more technically accurate way: 'conditions' vs 'ICD categories'. | No clinically significant difference |
3.2 | There would be no clinically significant difference in the safety and clinical performance of the device because it's the same population. The device says: "patients presenting visible skin structure abnormalities", whereas the equivalent says "patients who suffer from skin conditions". It's pretty much the same. | No clinically significant difference |
3.3 | There would be no clinically significant difference in the safety and clinical performance of the device because it's the same intended user. In both cases, the documentation says: "health care professionals (HCP)". The device is more specific in also mentioning that the technical team of the care provider is a relevant user during integration, but it's a higher level of specification, not a difference. | No clinically significant difference |
3.4 | There would be no clinically significant difference in the safety and clinical performance of the device because both show that there is enough evidence to establish the safety and performance of the devices when used in accordance with the IFU | No clinically significant difference |
Summary
It should come as no surprise that the analysis yelded the result of both devices being equivalent, because the device is an evolution of the equivalent device, which constitutes its previous generation.
The device is the result of focusing all the design efforts in the API, which constitutes the component that performs the clinical tasks also in the previous generation. Thanks to this, the new device is safer, more roboust and simpler than its predecesor. You can find more information about this on the following documents:
Legit.Health Plus description and specifications
R-TF-012-006 Lifecycle plan and report
Data generated and held by the manufacturer
Technical and pre-clinical data generated and held by us
In the clinical study LEGIT.HEALTH_DAO_Derivación_O_2022
conducted at Hospital Universitario de Cruces and four primary care centers, the primary care doctors exhibit low sensitivity, approximately 25% (but have high specificity at 96%). This low sensitivity might be due to the fact that all cases are referred to dermatologists. As a result, the primary care doctor's diagnosis doesn't impact the decision, making it a non-conservative diagnosis. If the diagnosis were made with the intention of either treating it in primary care or referring it to the dermatologist, it's likely that sensitivity would be significantly higher, while specificity would be lower.
This study reveals that approximately 29% of the referrals, even those from teledermatology, involve common and easily diagnosable conditions, with about half of them being related to seborrheic keratosis. Another example of conditions that can be confidently identified and managed without referrals includes skin tags, which the device can reliably confirm, and other entities are unlikely to misdiagnose. In the case of skin tags, the device correctly identifies them as the top1 ICD category with over 40% confidence.
The quality of the images significantly influences the performance of the system. This is a well-established fact in the field because image quality not only impacts the effectiveness of algorithms but also hinders dermatologists from making diagnoses through teledermatology systems. Specifically, poor-quality images of nevi often require an in-person consultation, causing unnecessary delays for specialists.
It's typically complex to calculate precise costs, but we can estimate that devices like ours could have a substantial impact on cost optimization while simultaneously reducing waiting times and expediting urgent cases.
In terms of the waiting list, the analysis assumes that patients could have received treatment earlier, and the appointment delays weren't due to personal reasons but rather a result of the hospital's waiting list.
Another study in Hospital Universitario de Cruces and Hospital Universitario de Basurto shows, again, strong evidence of the capacity of the device to assist healthcare practitioners in their clinical evaluations.
The device demonstrates great malignancy prediction and compelling image recognition capacity in the task of extracting ICD category features, such as such as melanoma, carcinoma, keratoses or nevi, with results similar to internal validation tests. The device achieved an excellent AUC of 0.8769 for the task of malignancy detection. In terms of generating a distribution of possible ICD categories, the device presented compelling top-3 and top-5 accuracies (72.85% and 81.64%), which confirms its utility as a tool to assist healthcare practitioners in a real clinical setting.
A third study held in Hospital Universitario de Torrejón, under the code LEGIT_COVIDX_EVCDAO_2022
, provides direct evidence based on the opinion of healthcare professionals that used the tool for a large period. We found out that the device proved to be helpful in their clinical practise. The main endpoint of the study, the Clinical Utility Questionnaire (CUQ), got very positive responses, indicating positive perceptions among specialists, particularly in terms of ease of use and effectiveness in optimizing consultation time according to each patient's needs.
Other secondary endpoints were also questionnaires that gathered relevant information such as the data utility and ease of use of the medical device. The Data Utility Questionnaire affirms unanimous agreement on the device's intended usefulness. System Usability Scale results showed high levels of user satisfaction and ease of navigation. Patient satisfaction scores reflected a positive overall experience with the tool. Importantly, no adverse events or reactions were observed, highlighting its favorable safety profile.
The device proves highly effective, safe, and user-friendly for managing dermatosis. Positive feedback from specialists and patients underscores its potential as a valuable clinical tool. The device exhibits significant clinical relevance for its intended use, offering objective clinical data for dermatosis evaluation.
Overall, the device proves to be a valuable support tool for healthcare practitioners and organizations in their clinical decision-making process.
Clinical data generated and held
We have disseminated our findings through four distinct articles, all accessible to the scientific community in reputable journals like JAAD and JID Innovations. These articles are:
- Automatic SCOring of Atopic Dermatitis using Deep Learning: A Pilot Study
- Automatic International Hidradenitis Suppurativa Severity Score System (AIHS4): A Novel Tool for Assessing Hidradenitis Suppurativa Severity with Artificial Intelligence
- Dermatology Image Quality Assessment (DIQA): Utilizing Artificial Intelligence to Ensure the Clinical Applicability of Remote Consultation and Clinical Trial Images
- Automatic Urticaria Activity Score (AUAS): Deep Learning-Based Automatic Hive Counting for Urticaria Severity Evaluation
These articles collectively demonstrate the performance of the medical device in the extraction of quantifiable data regarding the count, extent, and intensity of clinical signs, as well as evaluating the quality and clinical relevance of dermatological images.
In addition to our published literature, we have conducted additional and extensive analyses to assess the performance of these algorithms in extracting quantifiable clinical sign data and the distribution of potential ICD categories. Our Post-Market Clinical Follow-up activities have consistently shown that the device's performance aligns with or even surpasses the initial test results in specific scenarios.
In real-world applications, the device achieved results exceeding the specified metrics outlined in REQ_004_The user receives an interpretative distribution representation of possible ICD categories represented in the pixels of the image. Healthcare organizations that used the interpretative distribution representation of possible ICD categories for their clinical decision-making process as a main feature of the medical device, achieved impressive results, with 70% top1, 85% top3, and 92% top5 success rates.
When it comes to malignancy detection, the AUC (Area Under the Curve) obtained directly from user data is outstanding, with a value of 0.96. Even when focusing exclusively on specific ICD category groups such as the pigmented and tumoral dermatosis, a category where the proportion of malignant cases is higher than the overall average, the AUC remains excellent, consistently exceeding 0.90.
Clinical Data
Clinical Data From Literature
We have compiled all the clinical data from literature at the T-015-002 Pre-clinical and clinical evaluation record
.
In accordance with the T-015-001 Clinical Evaluation Plan
, the acceptance or rejection of each piece of information is determined through the application of the following weighted-based system:
Data characteristics | Methodological quality Q | Relevance, R | Contribution, C | Weighted value, W (W=Q+R+C) | Appraisal data | Notes |
---|---|---|---|---|---|---|
Very relevant information in relation with the product and its intended use | 30 | 30 | 30 | W ≥ 70 | Accepted | Pivotal data |
Relevant information in relation with the product and its intended use | 20 | 20 | 20 | 30 < W < 70 | Accepted | Other data |
Little relevant information in relation with the product and its intended use | 10 | 10 | 10 | W ≤ 30 | Rejected | No contribution, rejected |
As it is described at the T-015-001 Clinical Evaluation Plan
, preclinical data is accepted when it provides comparable technical information of equivalence/similar device on the market or information about similar accessories. On the other hand, preclinical data is rejected when the information is incomplete, insufficient, or irrelevant.
Scientific publications are accepted when the study design is adequate, and when they provide information that clarifies whether the use of the device is necessary, safe, and advantageous over other alternatives on the market. That is, when the intended use, the target patient population, or the target user group matches the device. On the other hand, scientific publications are rejected when their content do not discuss the topic of interest, when the conclusions are inconsistent or there are misinterpretations, when there is a lack of information on basic aspects, or when the information is inaccessible or repetitive.
Appraisal of clinical data
Skin structures analysis
In the literature review for skin analysis methods, 188 articles were compiled, of which 81 articles passed the Data Screening and 107 were discarded. These 81 articles that passed the Data Screening were reviewed thoroughly to assess their Quality, Relevance, and Contribution. As a result:
- 31 articles have been evaluated as highly relevant.
- 25 articles have been evaluated as relevant.
- 25 articles have been evaluated as non-relevant.
In this review, we identified different kinds of skin analysis methods that are intended for different use cases:
- Most commonly, these solutions are designed for skin-related image diagnosis (IDs 019, 020, 030, 034, 042, 044, 045, 047, 053, 058, 061, 063, 067, 068, 072, 078, 079, 080, 083, 085, 092, 096, 119, 122, 127, 128, 129, 132, 166), usually being able to detect 2 to 10 different skin diseases but in particular cases up to 174 different lesion categories. Of special relevance, certain solutions include novel techniques to increase the diagnosis performance. IDs 092 and 128 jointly process image and metadata information, ID 020 combines image and lesion segmentation information, ID 080 attaches a special device to a smartphone camera to improve the image quality and ease its processing, and IDs 042 and 132 include different tricks such as knowledge distillation, cost-sensitive learning, soft targets, cumulative learning or specific data augmentations to ease the learning with small and unbalanced datasets.
- Differently, other methods are designed for lesion segmentation with pixel-precision or for the severity assessment of different pathologies. In the case of lesion segmentation (IDs 016, 025, 027, 031, 037, 064, 076, 122), we find interesting novelties such as ID 016 that includes the use of special pre and post-processing steps, and ID 064 that proposes a new architecture designed with fewer parameters to perform more efficiently. And, as for the case of lesion severity assessment, we find some methods designed for the psoriasis (IDs 009) or acne (IDs 021, 035, 093) evaluation.
- Besides the design and development of these image analysis methods, we also find some other articles that include information related to their deployment and integration in phone apps and/or the cloud (IDs 019, 021, 035, 061, 087). Of special relevance, ID 021 proposes the pruning and feature-based knowledge distillation of the trained models to improve their efficiency in end-devices.
- Interestingly, we find also review articles that summarize the state-of-the-art of skin lesion analysis methods (IDs 001, 007, 009, 014, 015, 019, 113, 116, 120, 121, 146, 187, 188), including important tips for their design, deployment, and analysis of the challenges they face.
- Finally, other research articles also provide information related to the evaluation of these methods in clinical trials (IDs 019, 021, 030, 053, 081, 085, 140, 143, 184), providing meaningful information related to the performance of these methods in real environments and the benefits they offer to health professionals.
Facial palsy analysis
Regarding the literature review for facial palsy analysis methods, 46 articles were compiled, of which 31 articles passed the Data Screening and 15 were discarded. These 31 articles that passed the Data Screening were reviewed thoroughly to assess their Quality, Relevance, and Contribution. As a result:
- 15 articles have been evaluated as highly relevant.
- 10 articles have been evaluated as relevant.
- 6 articles have been evaluated as non-relevant.
In this review, we identified different articles that develop solutions intended for different use cases:
- Although we find some solutions focused on diagnosing facial palsy (ID 005) or predicting the probability of incurring it after specific surgeries (IDs 006, 009, 010), most works are based on the facial palsy severity assessment. To this end, these methods mainly rely on the analysis of facial landmarks (IDs 001, 003, 015, 016, 021, 022, 038), but we also find some other works that use manual labelling (ID 007), video or image processing (IDs 016, 018, 020), or automated emotion analysis (IDs 002, 013, 017, 019). For this severity assessment, some articles (IDs 003, 022, 046) also propose the creation of specific datasets to ease the analysis tasks.
- Interestingly, we find some article reviews (IDs 011, 012, 023, 024, 026) that describe the state-of-the-art of intelligent facial palsy-related methods.
- Finally, many of these solutions (IDs 001, 003, 005, 006, 007, 009, 010, 013, 014, 016, 017, 018, 019, 021) are involved in clinical trials to gather patient information for the method evaluation, providing meaningful information related to the performance of these methods in real environments and the benefits they offer to health professionals.
Clinical Data from Clinical Study Databases
The table shows the results obtained on the literature search in relation to similar devices used in the skin structures field. Regarding the clinical data found in relation to facial palsy, they use real patient data to train some model, but there are no specific metrics on whether they improve outcomes over doctors because in this case the tasks they do are different.
Name of similar device | Results discussed | References used to get the results |
---|---|---|
Fotofinder | Although their Moleanalyzer Pro tool just focused on moles, their findings suggest that dermatologists may improve their performance when they cooperate with the CNN (Convolutional Neural Network) and that a broader application of this human with machine approach could be beneficial for dermatologists and patients | 2804568 |
7-class skin disease recognition | This AI has been trained on skin-related data collected in hospitals from the Southwest of Ethiopia, Eastern Amhara, and Afar region. Final device works with an accuracy, precision, and sensitivity of above 97%, showing high safety and performance to be used as an assistive tool. | ID 019 |
Acne severity assessment app | The software developed shows a great performance for the acne severity assessment, being able to count and classify with high precision the different acne lesions. The performance of this app surpasses the General Practitioners' and gets close to the more experienced dermatologists. However, since the app could be biased towards the Chinese-like population, data from other regions should be included in the learning system. | ID 021 |
10-class cutaneous tumor recognition | This study shows how the use of this AI can assist dermatologists in increasing their lesion analysis performance. In particular, this boost of performance is bigger for the dermatologists with less experience | ID 030 |
mHealth app (CE-marked) | 40 skin diseases recognition AI trained in images of skin of colour from India. The app reflects an top-1 accuracy of 75% in clinical trials, 89% of top-3 accuracy, and 0.90 AU, showing its viability as a clinical decision support tool. A posterior independent study revealed that the app presents a sensitivity and specificity that surpasses the one from General Practitioners and gets close to the dermatologists' one. The performance also decreases depending on the phone device used, so the image quality should be taken into account. | ID 053 ID 081 |
174-class skin disease recognition | This study reveals that this AI model gets comparable top-1 accuracy (47.6%) to the dermatologists (49.7%) and residents (47.7%) but superior to the general practitioners (39.7%), showing promising capabilities as an assistance for skin lesion recognition | ID 085 |
Two applications able to recognize 47 lesion categories, that fulfill the CE-criteria, and registered as medical product at the Austrian Federal Office for Safety in Health Care | A clinical trial in Austria reveals that these two apps present high sensitivity and specificity (94-96%), probing its performance and suitability for skin lesion analysis assistance. | ID 140 |
Review of 272 clinical studies that include different AI solutions to facilitate the early diagnosis of skin cancers, especially in primary and community care settings. | This study reveals an average high accuracy for the recognition of melanoma (89%), squamous cell carcinoma (85%), basal cell carcinoma (87%), and malignancy estimation (88%). Although these numbers show the potential benefits of AI for skin lesion analysis (especially in primary care), some studies present different concerns related to the size, variability, and source of the studied population. | ID 143 |
Study of the benefits when using AI assistance for skin image analysis | The study involves 20 primary care physicians and 20 nurse practitioners with different levels of experience. When assisted by the AI, these practitioners increased their diagnosis agreement, demanded fewer biopsies and referrals, and increased their confidence. | ID 184 |
Clinical Data From Adverse Event Databases
We consulted the following sanitary alert databases specified at the table looking for incidents or alerts by using similar devices or methodologies as ours. We consulted in these databases the following aspects:
- The similar devices described: Dermengine, Fotofinder hadyscope pro app, Skinscreener, Skinvision and Triage.
- The following keywords: artificial intelligence, dermatology, deep learning, medical imaging, computer vision, as we had specified at the
R-TF-007-001 PMS plan 2023_001
.
Source of information | Link | Results | Analysis |
---|---|---|---|
FDA website MAUDE - manufacturer and User Facility Device Experience Searchable database | accessdata.fda.gov | 0 | |
FDA website Medical Device Recalls | accessdata.fda.gov | 0 | |
German Federal Institute for Drugs and Medical Devices (Bundesinstitut für Arzneimittel und Medizinprodukte, BfArM) | bfarm.de | 25 | Results obtained using the keywords "medical imaging". The incidences found have no relation to our device |
Swissmedic Swiss Competent Authority | swissmedic.ch | 71 | Results obtained using the keywords "medical imaging". The incidences found have no relation to our device, they are mainly related to X-ray and IVD devices |
AEMS Vigilancia de productos sanitarios | alertasps.aemps.es | 0 | |
MHRA Adverse events reporting | gov.uk | 15 | Results obtained using the keywords "artificial intelligence" and "deep learning". Although one result is related to a PACS, this one and the other incidences are not related to our device |
Ministero della Salute - Avvisi di sicurezza sui dispositivi medici | salute.gov.it | 1 | Result obtained using the keyword "dermatology", but the incidence was related to cosmetics. |
Regarding the similar devices, only the Triage search yielded results, that are not included within the table as they are related to cardiac devices.
Analysis of the Clinical Data
Requirement on safety
According to the T-015-001 Clinical Evaluation Plan
, we analyze the data accepted from the appraisal in order to reach conclusions about the safety and clinical performance of the device including its clinical benefits and possible adverse events, following stipulations from Appendix A7.1 of MEDDEV 2.7.1. rev.4.
.
Special design features and safety concerns
The device leverages advanced computer vision algorithms to analyze images of the skin and its various structures. There are no medicinal, human, or animal components present in the device, mitigating concerns related to biocompatibility and potential adverse reactions to biological materials. However, the risk management documentation identified potential safety concerns related to the accuracy and reliability of the device, particularly in cases of poor image quality or uncommon skin conditions.
Addressing risks from risk management documentation and literature
All risks identified in the risk management documentation have undergone thorough evaluation from both a technical and clinical perspective. This includes ensuring the device's performance remains robust across diverse patient populations, skin types, and conditions. Clinical precautions for the reduction of risks and the management of potential adverse events have been outlined, with particular attention paid to ensuring healthcare practitioners are aware of the device's role as a supplementary tool in clinical decision-making, rather than a standalone diagnostic solution.
Training and user precautions
Given the sophisticated nature of the technology, there is a requisite need for user training to ensure the device is used correctly and effectively. This training will cover proper image capture techniques, interpretation of the device's output, and understanding the device's limitations. The Intended users, healthcare organizations, and their stakeholders, including healthcare practitioners, have been deemed adequate, given their professional background and experience in clinical settings. All training requirements and user precautions have been explicitly described in the Instructions for Use (IFU) accompanying the device.
Consistency with current knowledge and state of the art
There is full consistency between the current knowledge/state of the art in computer vision applications in dermatology, the clinical data available for the device, the information materials provided by the manufacturer, and the device's risk management documentation. The clinical benefits, as well as any potential risks associated with the device's use, are clearly communicated to the end-users, ensuring transparency and promoting safe usage.
Cybersecurity
The cybersecurity framework for the device aims to ensure compliance with EU regulations while following international guidelines and best practices. This strategy is grounded in recognized standards such as MDCG 2019-16 and IMDRF/CYBER WG/N60FINAL:2020. A comprehensive approach to managing cybersecurity risks is implemented, beginning with an initial risk assessment, which overlaps with the broader risk management process. This ensures that all identified cybersecurity risks are integrated into the device's lifecycle risk management strategy, evolving continuously to adapt to new threats.
One of the key aspects of cybersecurity management involves identifying potential risks to the device's AI/ML models and auxiliary software components. Threats may arise from unauthorized access, data manipulation, or exploitation of vulnerabilities within the AI/ML models themselves. Risk assessments focus on evaluating various attack vectors, including network vulnerabilities, physical access threats, and potential weaknesses introduced by third-party integrations.
Cybersecurity risk management must be harmonized with safety risk management to ensure a balanced approach that protects both device functionality and patient safety. After identifying potential risks, their impact on patient safety, data integrity, device reliability, and overall functionality is evaluated. This evaluation enables the prioritization of risks, focusing on those with the highest potential impact.
To mitigate identified risks, a range of protective measures are implemented. These include technical solutions such as encryption and access controls, as well as organizational strategies like staff training and policy development. The goal is to develop robust security features embedded within the AI/ML models and their auxiliary components, without compromising the device's functionality.
Security by Design principles are applied from the earliest stages of development, ensuring that cybersecurity is an integral component throughout the lifecycle of the AI/ML models and their auxiliary software. Security controls like user authentication, data encryption, and secure coding practices are implemented to protect the AI/ML models from unauthorized access and data breaches.
Data protection and privacy are also central to the cybersecurity strategy, particularly regarding the handling and storage of sensitive patient information. Techniques such as data anonymization help ensure compliance with regulatory requirements, safeguarding both patient privacy and data integrity.
A robust incident response plan is established to address potential security breaches, outlining the necessary steps for containment, investigation, and recovery.
Finally, the cybersecurity measures are reflected in the device's Instructions for Use (IFU), ensuring transparency and ease of access for users. The IFU is maintained with stringent security practices, incorporating version control and secure development methods to guarantee its integrity. Through regular updates and clear documentation, users are kept informed of all relevant cybersecurity features and protections.
Clinical performance
The aim of the clinical performance is to demonstrate its capacity to generate clinically relevant outputs aligned with its intended purpose of supporting dermatological assessments through image analysis. This involves validating that the software provides outputs that positively impact individual health by contributing to measurable, patient-relevant outcomes. Specifically, the device aims to aid healthcare professionals in diagnosing and monitoring skin abnormalities, improving diagnostic accuracy and patient management. By providing reliable and actionable insights that facilitate risk prediction, screening, and clinical evaluations, the device seeks to play a valuable role in enhancing patient care and public health within professional healthcare settings.
Specific performance and safety questions
The clinical performance evaluation for the device aims to answer the following questions:
- Does the device effectively assist healthcare professionals in evaluating dermatological conditions by analyzing images of the skin structures, providing data that supports clinical decision-making?
- Does the device consistently support the monitoring of skin conditions over time, enabling accurate assessment for healthcare providers?
- Does the device maintain the quality and accuracy of its image analysis and data processing to ensure reliable support for clinical evaluations?
- Does the device exhibit an acceptable safety profile, with a risk level comparable to or lower than other AI-based dermatological assessment tools commonly used in clinical practice?
- Does the device demonstrate equivalent or superior performance compared to other dermatological analysis tools in providing reliable clinical data to support patient evaluations?
These clinical questions aim to assess the effectiveness, safety, and performance of the device in contexts where healthcare practitioners can leverage the device's image analysis to support evaluations of skin conditions, ensuring it meets the needs of healthcare professionals and patients while maintaining regulatory compliance and safety standards.
Type of clinical performance evaluation
The clinical performance for the device is based on the available post-market clinical investigations pertinent to the device under evaluation and the scientific literature.
This evaluation includes the different data sources described in the section below (Identification of relevant data
).
Identification of relevant data
Data relevant to the clinical evaluation of the device is listed in the table below.
Data Source | Section |
---|---|
Data from State of the art (SOTA) | Section Clinical background and SOTA analysis |
Data generated and held by the manufacturer | Section Data generated and held by the manufacturer |
Risk Management and Risk Analysis | Section Risk Management and Risk Analysis |
Data from Post-Market Surveillance | Section Data from Post-Market Surveillance (PMS) |
Data from Post-market clinical investigations | Section Post-market Clinical investigations |
Data from vigilance reporting | Section Data from Vigilance |
Data retrieved from the literature | Section Data from literature review |
Data generated and held by the manufacturer
As indicated in the CEP for the device, the present clinical evaluation shall also consider the data generated and held by the manufacturer, which includes all clinical data generated from risk management activities.
Risk Management and Risk Analysis
We have established, implemented, documented, and maintained a risk management system in accordance with Annex I of MDR. This involves the establishment of a risk management plan for The device, allowing traceability for every hazard, identification and evaluation of risks, accomplishment and verification of the risk control measures, and verification that the residual risks are accepted. Risk management activities for the device have been performed in accordance with ISO 14971:2019.
Regarding the application of the device by IT professionals and healthcare professionals, appropriate warnings and precautions are listed in the IFU. The IFU identifies the contra-indications of the product.
The risk analysis also covers all the life-cycle phases of the device. The risk management assessment of The device indicates an overall acceptable benefit-risk ratio. The primary benefits include improved precision in analyzing skin structures, which enhances clinical outcomes without invasive procedures. The device also supports healthcare workflow efficiency by integrating seamlessly into clinical settings, enabling reduced patient wait times and optimized resource use. Its capacity for monitoring provides personalized patient care and valuable data for extended clinical studies. Additionally, the device aids preliminary diagnosis by offering interpretative representations of possible ICD classifications.
Key risks identified include data privacy and security concerns, the risk of healthcare practitioners becoming overly reliant on the device, and potential data misinterpretation. To address these, the device complies with FHIR standards, and the Instructions for Use (IFU) provide guidance to mitigate interpretation risks. Continuous monitoring during post-market surveillance is recommended to address any residual risks related to data misinterpretation.
The benefit-risk acceptability relies on demonstrated clinical advantages, efficiency, and minimal associated risks. The device is designed for a broad patient population and meets unmet needs in dermatological assessment. The residual risk remains low, provided that data privacy, proper usage, and ongoing surveillance are upheld, ensuring the device's continued safe and effective operation in clinical environments.
As indicated in the risk management report (R-TF-013-003
), after all risk control measures have been implemented and verified, and benefit/risk analysis has been performed for each risk and for the overall residual risk, it is concluded that the overall residual risk of the device is acceptable.
The potential risks which may be associated with the use are:
Hazard ID | Hazard | Mitigation measures | Severity | Likelihood | Residual risk evaluation |
---|---|---|---|---|---|
5 | Incorrect clinical information | Information about device outputs and intended user (HCP) are detailed in the IFU. The medical device returns metadata about the output that helps supervising it, such as explainability media and other metrics. | 3 | 2 | As far as possible |
6 | Incorrect diagnosis or follow up | Information about device outputs and intended user (HCP) are detailed in the IFU. The medical device returns metadata about the output that helps supervising it, such as explainability media and other metrics The device returns an interpretative distribution representation of possible ICD categories, not just one single condition. | 3 | 2 | As far as possible |
9 | Image artefacts/resolution | A requirement of the device defines the creation of a processor whose purpose is to ensure that the image have enough quality. In other words, an algorithm, similar to the ones used to classify diseases, is used to check the validity of the image and provides an image quality score. The device returns meaningful messages to the users about the quality score of the images, this allows care providers to re-take a photo. We also offer training to the users to optimize the imaging process so that it is optimal for the device's operation. | 3 | 2 | As far as possible |
11 | Data transmission failure from healthcare provider's system | State-of-the-art techniques of security and software availability. The device returns meaningful messages about the error to help troubleshooting. | 3 | 2 | As far as possible |
12 | Data input failure | State-of-the-art techniques of security and software availability. The device returns meaningful messages about the error to help troubleshooting. | 3 | 2 | As far as possible |
13 | Data accessibility failure | State-of-the-art techniques of security and software availability. The device returns meaningful messages about the error to help troubleshooting. | 3 | 2 | As far as possible |
14 | Data transmission failure | State-of-the-art techniques of security and software availability. The device returns meaningful messages about the error to help troubleshooting. The endpoints of the device follow HL7's FHIR interoperability standard. | 3 | 2 | As far as possible |
30 | Inadequate lighting conditions during image capture | A requirement of the device defines the creation of a processor whose purpose is to ensure that the image have enough quality. In other words, an algorithm, similar to the ones used to classify diseases, is used to check the validity of the image and provides an image quality score. The device returns meaningful messages to the users about the quality score of the images, this allows care providers to re-take a photo. We also offer training to the users to optimize the imaging process so that it is optimal for the device's operation. | 3 | 2 | As far as possible |
Sources: R-TF-013-002 Risk Management Record, R-TF-013-003 Risk Management Report
Data from Post-Market Surveillance (PMS)
Data from post-market surveillance report (PMS report)
The last PMS report (R-TF-007-004
) was released on 2024. This PMS report relied on data from the previous generation of the device (Legit.Health). The main conclusions from the PMS report are summarized below:
- Sales: The previous generation was first available by the end of 2020. Since then, 21 contracts were signed with 21 customers, ranging from remote to on-site use of the device, and from government-run care providers to for-profit care providers. During this period, more than 4500 reports have been created by more than 500 practitioners who have used the product to help more than 1000 patients.
- Serious Incidents & Field Safety Corrective Actions (FSCA): No serious incidents or FSCA were reported for the device's previous generation during the reviewed period.
- Non-serious incidents: in the latest PMS report, 7 non-serious incidents have been identified (4 coming from customer complaints and 3 coming from internal nonconformities). All the 7 non-serious incidents have been documented, evaluated and addressed. The non-serious incidents were mainly due to slow response of the API, inaccuracy in the probability distribution of ICD categories.
- Trend Reports: No significant trends or deviations requiring actions were observed in the device's safety and performance, as the focus remained on the previous generation of the device. The PMCF evaluation highlighted consistent positive feedback on user satisfaction, with both primary and secondary healthcare professionals reporting favorable experiences.
- Post-Market Clinical Follow-Up (PMCF) Plan: Activity 2 provided valuable insights into device performance, although no clear trends emerged.
- Corrective and Preventive Actions (CAPA): 30 CAPA have been initiated and they are distributed as follows:
- all CAPA are classified as corrective action
- 22 CAPA are related to non-conformity in our processes, 8 CAPA are related to non-conformity in our device
- 3 CAPA were triggered by customer complaints, 6 CAPA were triggered by internal non-conformities, 21 CAPA were triggered by audit non-conformities (both client and internal audits)
- 2 CAPA are under final implementation, 3 CAPA are under evaluation to verify their effectiveness, 25 CAPA are closed with effectiveness verified
- none of the CAPA includes the initiation of FSCA.
Conclusion
While no serious incidents or significant safety concerns were identified, various corrective actions were taken to improve the system's performance, including refining image analysis processes and addressing service-related issues.
Data from Post-Market Clinical Follow-up (PMCF) evaluation report
The last PMCF evaluation report (R-TF-007-005
) was reviewed in 2024.
The PMCF report conclusions highlight the following key findings:
- Clinical Literature Review: The review of state-of-the-art publications has validated the device's safety for its intended use, ensuring it aligns with current scientific and clinical knowledge.
- PMCF Studies: These studies have confirmed the safety, performance, and clinical benefits of the device in real-world environments, identifying new risks or adverse events, and providing additional clinical evidence to support its ongoing use.
- Image Recognition Processor Success Metrics: Performance analysis through success metrics has validated the device's operational efficacy, with reports from patients and healthcare practitioners confirming it functions as intended.
- Similar Devices Comparison: Comparing the device with similar products on the market and consulting sanitary alert databases has helped assess its position and safety profile, ensuring proactive risk management.
- Feedback and Complaint Analysis: User feedback and complaint analysis have demonstrated high user satisfaction, with successful CAPAs addressing complaints and contributing to device improvement.
- Integration into Clinical Evaluation and Risk Management: PMCF findings will inform ongoing clinical evaluation and risk management, ensuring the device's safety and efficacy are maintained.
- Preventive and Corrective Measures: Identifying non-conformities and resolving them through CAPAs emphasizes continuous improvement, contributing to the device's overall enhancement.
These activities collectively support the device's safe and effective use post-market, with ongoing efforts to refine and improve its performance.
Post-market Clinical investigations
The completed and on-going post-market clinical investigations performed for the device under evaluation are summarized in the table below.
Title | Code | Endpoints (Acceptance Criteria) | What have we achieved? |
---|---|---|---|
Clinical validation study of a CAD system with artificial intelligence algorithms for early noninvasive detection of in vivo cutaneous melanoma. | LEGIT_MC_EVCDAO_2019 | AUC greater than 0.8 Sensitivity of 80% or higher Specificity of 70% or higher | We get an AUC of 0.842 identifying melanoma. An AUC of 0.8983 detecting malignancy. A top-3 sensitivity of 0.9032.[1] A top-1 specificity of 0.8054. |
Clinical Validation of a Computer-Aided Diagnosis (CAD) System Utilizing Artificial Intelligence Algorithms for Continuous and Remote Monitoring of Patient Condition Severity in an Objective and Stable Manner. | LEGIT_COVIDX_EVCDAO_2022 | A score of 8 or higher in the Clinical Utility Questionnaire (CUS) filled by medical staff. | At the end of the study we got a mean of 76.67 (7.667) in the CUS. It was a positive feedback, although it wasn't the pretended one. It has been discussed in the discussion section due to the small sample size of the study. |
Optimization of clinical flow in patients with dermatological conditions using Artificial Intelligence. | LEGIT.HEALTH_IDEI_2023 | An improvement of diagnostic accuracy of 10% (Ferri et al. 2020) Scores equal or greater than 70 on the System Usability Scale (SUS) An AUC equal or greater than 0.8 detecting malignancy A sensitivity of 80% detecting malignancyA specificity of 70% detecting malignancy | We have concluded the first part of the study, the second part will start in Q1, 2025. The results are as follow: Retrospective analysis Detecting malignancy: dermatologists and the medical device achieved an AUC of 0.79 and 0.76 respectively. Sensitivity: Dermatologists 86%, device 81%. Specificity: Dermatologists 36%, Device 52%. Diagnosis accuracy: Dermatologists Top-1 0.56, top-3 0.70; device: top-1 0.50, top-3 0.71, top-5 0.78. Prospective analysis Detecting malignancy: Dermatologists + Device: AUC 0.94, sensitivity 88%, specificity 85%. Accuracy on diagnosis: Dermatologist + device: top-1 0.85. |
Non-Invasive Prospective Pilot in a Live Environment for the improvement of diagnosis of skin pathologies in primary care and dermatology | LEGIT.HEALTH_SAN_2024 | An improvement of diagnostic accuracy on both primary care physicians and dermatologists A positive view of Legit.Health regarding diagnosis support A reduction of 30% of referral to dermatology (Warshaw et al. 2011) An improvement in remote consultations | An increase of diagnosis accuracy of 27.02% in primary care and 10.46% in dermatology. 60.89% of consultations should not refer according to primary care and 53.59% according to dermatology. Regarding remote consultations: primary care physicians 59,68% can be handed remotely; for dermatologists 47,71%. The design and usability of Legit.Health received an average score of 8, with all respondents giving it the same rating, indicating strong consensus. |
Non-Invasive Prospective Pilot in a Live Environment for the improvement of diagnosis of Generalized Pustular Psoriasis | LEGIT.HEALTH_BI_2024 | An improvement of diagnosis accuracy of generalized pustular psoriasis (GPP). An improvement of diagnosis accuracy of other skin conditions such as hidradenitis suppurativa or palmoplantar pustulosis on both, primary care physicians and dermatologists. | An increase of diagnosis accuracy of 17.42% in primary care and 8.40% in dermatology.An increase of diagnosis accuracy of 22.97% in generalized pustular psoriasis for both specialities.An increase of 8.92% of the accuracy on the diagnosis of hidradenitis suppurativa.An increase of 34.38% of the accuracy on the diagnosis of palmoplantar pustulosis. |
Non-Invasive Prospective Pilot in a Live Environment for the improvement of the diagnosis of skin pathologies in primary care | LEGIT.HEALTH_PH_2024_NIPPLE | An improvement of diagnostic accuracy on both primary care physicians. A positive view of Legit.Health regarding diagnosis support. A reduction of 30% referral to dermatology. An improvement in remote consultations. | An increase of diagnosis accuracy of 9.29% in primary care. 48.89% of cases did not necessitate a referral. The results show that 60.74% of the cases can be handled remotely. |
Pilot study for the clinical validation of an artificial intelligence algorithm to optimize the appropriateness of dermatology referrals.(Ongoing Study) | LEGIT.HEALTH_DAO_Derivación_O_2022 | Improve the adequacy of referrals to dermatology A reduction of waiting lists (at least 30% Warshaw et al. 2011) A reduction of the costs in secondary care | We have currently recruited 79 patients. Study is ongoing and recruiting patients. |
[1] The top-1
, top-3
, or top-5
terminology refers to the output generated by the medical device when analyzing an image. The device provides a list of the five most likely diagnoses based on the analysis. Top-1
refers to the first diagnosis, top-3
means the condition (in this study, melanoma) is among the three most likely diagnoses, and top-5
means it is included within the five most probable diagnoses. In this context, since the device is intended as a diagnostic aid rather than a diagnostic tool itself, the most critical aspect is that the target condition appears within the top-5 diagnoses, which helps confirm the physician's diagnosis.
Data from Vigilance
A search for safety incidents and alerts was conducted for the similar products, which have been described in section Similar devices
(SkinVision and MoleScope).
The information on vigilance was obtained from the following vigilance databases:
- FDA (Food and Drug Administration): Provides Total Product Lifecycle data, including enforcement reports, warning letters, MAUDE database reports, CDRH inspections database, FDA recall database, and TPLC database.
- Swissmedic: Swiss agency for the authorization and supervision of therapeutic products, offering a recall list of medical devices within the scope of market surveillance.
- MHRA (Medicines and Healthcare Products Regulatory Agency): An executive agency of the Department of Health in Great Britain, responsible for ensuring the effectiveness and safety of medicines and medical devices.
- AEMPS Vigilancia de productos sanitarios: A state agency in Spain attached to the Ministry of Health, responsible for guaranteeing the quality, safety, efficacy, and accurate information of medicines and health products.
Since the device is not introduced to the European market, the adverse events for similar devices were investigated evaluating PMS complaints in various international regulatory databases.
The results are summarized in the following table.
Source of information | Link | Search Team | Alerts | Relevant | Data Range |
---|---|---|---|---|---|
FDA websiteEnforcement ReportSearchable database | http://www.fda.gov/Safety/Recalls/EnforcementReports/default.htm | MoleScopeSkinVision | 0 | N/A | N/A |
FDA websiteWarning lettersSearchable database | http://www.fda.gov/ICECI/EnforcementActions/WarningLetters/default.htm#recent | MoleScopeSkinVision | 0 | N/A | N/A |
FDA websiteMAUDE - manufacturer and User Facility Device ExperienceSearchable database | http://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfMAUDE/search.CFM | MoleScopeSkinVision | 0 | N/A | 2014-11-04 / 2024-11-04 |
FDA websiteMedical Device Recalls | http://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfRes/textsearch.cfm | MoleScopeSkinVision | 0 | N/A | N/A |
FDA websiteTPLC - Total Product Life Cycle database | http://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfTPLC/tplc.cfm | MoleScopeSkinVision | 0 | N/A | Since 2014 |
SwissmedicSwiss Competent Authority | https://www.swissmedic.ch/swissmedic/en/home/medical-devices/fsca.html | MoleScopeSkinVision | 0 | N/A | 2014-11-04 / 2024-11-04 |
AEPMS Vigilancia de productos sanitarios | https://alertasps.aemps.es/alertasps/alertas | MoleScopeSkinVision | 0 | N/A | 2014-11-04 / 2024-11-04 |
MHRA Adverse events reporting | https://www.gov.uk/drug-device-alerts | MoleScopeSkinVision | 0 | N/A | 2014-11-04 / 2024-11-04 |
Vigilance conclusions
A vigilance search was conducted to identify any safety incidents and alerts related to products similar to the device, specifically SkinVision and MoleScope. This search involved several key vigilance databases, including the FDA (with data from sources like the MAUDE database, TPLC database, and FDA recall and warning letter databases), Swissmedic (for Swiss market surveillance recall data), MHRA (UK's Medicines and Healthcare Products Regulatory Agency for device alerts), and AEMPS (Spain's health product safety agency).
No relevant safety incidents, alerts, or adverse events were found across all databases for MoleScope and SkinVision within the search period, spanning from November 2014 to November 2024. This indicates that, to date, similar products have not raised notable safety concerns in international regulatory databases, which supports the anticipated vigilance profile for the device.
Data from literature review
As part of the clinical evaluation, a search, compilation and assessment of clinical data relevant to this device has been carried out, as set out in the CEP (R-TF-015-001 Clinical Evaluation Plan
).
Identification of relevant bibliographic data
A comprehensive search of selected scientific literature databases (PubMed (MEDLINE)) was carried out on 2024-10-29.
The literature searches were performed according to the criteria defined in the Clinical Evaluation Plan. The algorithms used were the following:
- PubMed:
("skin cancer" OR "epidermis" OR "chronic skin conditions" OR "skin conditions" OR "inflammatory skin diseases" OR "malignant skin lesions" OR "melanoma" OR "acne" OR "psoriasis" OR "dermatofibroma" OR "dermatosis") AND ("Legit.Health" OR "software" OR "digital imag\*") AND ("SkinVision" OR "artificial intelligence" OR "machine learning" OR "computer vision" OR "smartphone") AND ("performanc\*" OR "safe\*" OR "clinical")
The searches were performed for a period of 10 years (2014-10-29 to 2024-10-29).
The search process is summarised below:
- Results of the search in PubMed: 61 articles
The 61 articles are listed in Attachment 02: Literature Search Records in the Clinical Performance folder.
A bibliographic screening of articles that are not related to the medical condition and the device under evaluation (software devices for the assessment of skin structures) was performed and 49 articles were discarded. Articles for which the full content is not accessible, which are not in English and which are not related to the product under evaluation or the medical condition were discarded.
The remaining 12 articles were appraised according to the appraisal criteria defined in CEP and 6 articles, whose score was equal or lower than 10 points were discarded.
The literature review is summarized below:
- Title and abstract review: 61 articles
- Screening: 49 articles
- Appraised articles: 12 articles
- Exclusion based on appraisal criteria defined in CEP: 6 articles
- Total studies included in the review: 6 articles
Evaluation of clinical data
To determine the value of the data identified in the literature, the results obtained from the search in PubMed were screened on the basis of the title and the information contained in the abstract.
The assessment of the relevant data has been carried out according to the assessment plan described in the CEP (R-TF-015-001 Clinical Evaluation Plan
).
The 49 excluded articles were discarded mainly because their content is not related to the device or the medical conditions according to its intended use. The remaining 12 articles were then evaluated, and 6 articles were finally selected for analysis. The detailed justification for the exclusion of the discarded articles can be found in Attachment 02: Literature Search Records.
Product Specific Data - Summary and Evaluation of Clinical Data
The 12 articles selected for their appraisal were evaluated and scored according to the appraisal criteria defined in CEP. Finally, 6 articles were selected and analyzed in the following section.
Analysis of clinical data
The following is a summary of the four articles selected for evaluation.
Article ID 1
- Title / Year: Implementation of artificial intelligence for the detection of cutaneous melanoma within a primary care setting: prevalence and types of skin cancer in outdoor enthusiasts (2023)
- Author: Miller, IJ
- DOI / PMID: 10.7717/peerj.15737
- Brief Summary of the article: The article discusses the use of artificial intelligence (AI) and machine learning, particularly convolutional neural networks (CNN), as tools in skin cancer screening. AI technology, including teledermatology and high-definition dermatoscopy, is increasingly applied in primary care for non-invasive, early skin cancer detection. AI analysis of individual skin lesions, using datasets of reference images, has shown high sensitivity in distinguishing melanomas from benign lesions, with some algorithms demonstrating sensitivity similar to that of trained dermatologists. For instance, the Fotofinder Moleanalyzer achieved sensitivity and specificity rates comparable to dermatologists, proving effective in identifying melanomas. Studies on CNN-based AI systems reveal that they can enhance dermatologists' diagnostic accuracy, with one example showing an increase from 60% to 75% in sensitivity and from 65% to 73% in specificity when AI-assisted. Another AI algorithm reached 100% sensitivity and a high receiver operating characteristic (91.8%) in detecting suspicious skin lesions from a large image set, outperforming dermatologists in some cases. However, the article notes a gap in real-world data, as most studies rely on datasets rather than clinical settings. The study's findings underscore the potential of AI-based imaging to aid clinical decision-making in identifying high-risk lesions, though limitations in real-world implementation remain.
- Risks Interpretation:
- Data Limitations and Bias: Many AI models are trained on limited datasets that may not represent the full diversity of skin types, lesion types, or conditions encountered in clinical settings. This can lead to biases in AI assessments, potentially causing inaccuracies, particularly with underrepresented skin types.
- False Positives and False Negatives: Although AI can be highly sensitive, it may yield false-positive results (incorrectly identifying benign lesions as suspicious) or false negatives (failing to detect true melanomas). These misclassifications could lead to unnecessary anxiety, unwarranted biopsies, or delayed diagnosis.
- Dependency on Image Quality: The accuracy of AI assessments is significantly affected by the quality and consistency of images. Poor lighting, resolution, and angle may compromise results, especially in teledermatology settings, leading to potential misdiagnoses.
- Over-Reliance on AI by Non-Specialists: While AI tools are increasingly accessible to primary care providers, there is a risk that less experienced clinicians may over-rely on AI results. This could reduce diagnostic vigilance, with clinicians relying on the tool without sufficient expertise to verify its findings.
- Challenges in Real-World Implementation: Many AI models show promising results in controlled settings but may struggle with performance and reliability when applied in diverse real-world environments. The lack of clinical validation data for AI models complicates their integration into healthcare workflows, which could introduce risks if not thoroughly vetted.
- Liability and Accountability: The article also notes the unclear regulatory and legal frameworks surrounding AI, raising questions about accountability in cases of misdiagnosis. If an AI tool contributes to an incorrect diagnosis or treatment decision, responsibility is often ambiguous, which complicates both clinical practice and patient safety.
- IFUs claims interpretation:
- Intended Purpose: The device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs, and interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: The device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus, and nails) and associated mucous membranes (conjunctival, oral, and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
- Contraindications: We advise not to use the device in: Skin structures located at a distance greater than 1 cm from the eye, beyond the optimal range for examination. Skin areas that are obscured from view, situated within skin folds or concealed in other manners, making them inaccessible for camera examination. Regions of the skin showcasing scars or fibrosis, indicative of past injuries or trauma. Skin structures exhibiting extensive damage, characterized by severe ulcerations or active bleeding. Skin structures contaminated with foreign substances, including but not limited to tattoos and creams. Skin structures situated at anatomically special sites, such as underneath the nails, requiring special attention. Portions of skin that are densely covered with hair, potentially obstructing the view and hindering examination.
Article ID 31
- Title / Year: AI outperformed every dermatologist in dermoscopic melanoma diagnosis, using an optimized deep-CNN architecture with custom mini-batch logic and loss function (2021)
- Author: Pham, TC
- DOI / PMID: 10.1038/s41598-021-96707-8
- Brief Summary of the article: This study addresses the use of AI, specifically deep convolutional neural networks (CNNs), to detect melanoma from skin lesion images. With melanoma being a highly deadly skin cancer, early detection is crucial, and recent advancements in computer vision within AI have shown potential for automating its diagnosis through image analysis. The study identifies challenges, such as data limitations and class imbalances, with melanoma images being much fewer than non-melanoma ones in datasets. This imbalance impacts the model's sensitivity to melanoma (minority class) versus nevus (majority class). To address this, the researchers propose a customized AI model with a custom loss function (CLF) and custom mini-batch logic that specifically adjusts training to better detect melanoma in imbalanced datasets. The model uses real-time image augmentation and specialized fully connected layers optimized for binary melanoma classification, achieving a high AUC, sensitivity (SEN), and specificity (SPE). Among tested CNN architectures, DenseNet169 performs best due to its dense connections, achieving superior sensitivity and specificity compared to dermatologists on benchmark datasets. The study's proposed techniques could extend to other medical image classifications, demonstrating the potential for customized AI to enhance diagnostic accuracy in clinical applications.
- Risks Interpretation:
- Data Imbalance Risks: With fewer melanoma images compared to non-melanoma in available datasets, there is a significant risk of the model underperforming in detecting melanoma cases. This imbalance may lead to reduced sensitivity (SEN) toward melanoma, meaning the model could miss critical cases, potentially resulting in delayed diagnosis or misdiagnosis of patients.
- Overfitting and Generalization Issues: The limited number of annotated melanoma images raises concerns about overfitting, where the model performs well on training data but poorly on new data. This could result in unreliable predictions when the model encounters real-world, varied patient images, limiting its clinical utility.
- Misclassification Costs: The cost of misclassifying melanoma as a non-melanoma condition is much higher than vice versa, as failing to detect melanoma early can be life-threatening. However, the article notes that it's challenging to quantify these costs in training the AI model, increasing the risk of inappropriate weighting and potential diagnostic errors.
- Model Dependence on Dataset Quality: The performance of AI models heavily relies on high-quality, representative datasets. If the datasets lack diversity or contain inaccuracies, the model's predictions may be unreliable, posing risks to patient safety.
- Technical Complexity of Solutions: Custom loss functions and mini-batch logic are complex to implement and may be less adaptable across other medical image classification tasks. This could limit scalability and complicate model validation, increasing the risk that the solution may not be broadly applicable or easy to update.
- IFUs claims interpretation:
- Intended Purpose: The device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing quantification of intensity, count, extent of visible clinical signs, and interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: The device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails), and associated mucous membranes (conjunctival, oral, and genital), the dermis, the cutaneous vasculature, and the subcutaneous tissue (subcutis).
- Intended User: The software as a medical device is specifically intended for use by IT professionals (ITPs) working within healthcare organizations who are responsible for integrating the software into the healthcare system infrastructure.
- Contraindications: We advise not to use the device in: Skin structures located at a distance greater than 1 cm from the eye, beyond the optimal range for examination. Skin areas that are obscured from view, situated within skin folds or concealed in other manners, making them inaccessible for camera examination. Regions of the skin showcasing scars or fibrosis, indicative of past injuries or trauma. Skin structures exhibiting extensive damage, characterized by severe ulcerations or active bleeding. Skin structures contaminated with foreign substances, including but not limited to tattoos and creams. Skin structures situated at anatomically special sites, such as underneath the nails, requiring special attention. Portions of skin that are densely covered with hair, potentially obstructing the view and hindering examination.
- Use Environment: The device is intended to be used in the setting of healthcare organizations and their IT departments, which commonly are situated inside hospitals or other clinical facilities.
- Operating Principle: The device is a computational medical tool leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures.
- Part of the Body: The device is intended to analyze images of the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus and nails), and associated mucous membranes (conjunctival, oral, and genital), the dermis, the cutaneous vasculature, and the subcutaneous tissue (subcutis). Due to the nature of the device (stand-alone software), it doesn't come into contact with tissue or bodily fluids.
Article ID 34
- Title / Year: Artificial intelligence-based image classification methods for diagnosis of skin cancer: Challenges and opportunities (2020)
- Author: Goyal, M.
- DOI / PMID: 10.1016/j.compbiomed.2020.104065
- Brief Summary of the article: The use of artificial intelligence (AI), particularly convolutional neural networks (CNNs), is advancing in dermatology for assessing skin conditions, primarily focused on skin cancer detection. Traditional diagnostic practices rely on visual examination by dermatologists, often enhanced by dermoscopy and biopsy, but AI is poised to transform this process by enabling automated, computer-aided diagnostics (CAD) through medical imaging. Algorithms have shown promising results, frequently matching or surpassing clinicians in classifying lesions as malignant or benign based on large datasets of clinical, dermoscopic, and histopathology images. Esteva et al.'s deep learning model, for example, achieved performance on par with dermatologists for differentiating between melanoma and benign lesions. Multiple public datasets, such as the ISIC Archive, Interactive Atlas of Dermoscopy, and HAM10000, support these AI-driven advancements by providing annotated images for algorithm training. AI solutions apply to various imaging modalities like dermoscopy, clinical photos, and whole-slide pathology scanning, adapting to high-resolution images captured through devices ranging from DSLR cameras to mobile phones. Studies by Codella, Haenssle, and Brinker illustrate that ensemble and deep learning methods can perform comparably or even better than dermatologists in specific cases. Challenges persist, however, as AI algorithms depend on balanced, high-quality datasets for accuracy and face limitations in real-world settings due to patient-specific factors like skin type and lifestyle, which these models currently overlook. Further collaboration between AI and clinical fields is essential to refine these tools for consistent, accessible, and cost-effective skin cancer diagnostics.
- Risks Interpretation:
- Data Quality and Diversity: AI algorithms require large, diverse, and balanced datasets to achieve high diagnostic accuracy. Limited or unbalanced data can result in overfitting, bias, or reduced performance on less-represented skin types, ethnicities, or lesion types, leading to potential misdiagnoses.
- Intra- and Inter-Class Variability: Skin lesions can show significant intra-class similarity (e.g., similar benign lesions may appear different under varying conditions) and inter-class dissimilarity (e.g., two benign and malignant lesions may look alike), complicating accurate classification by AI models. This variability increases the risk of diagnostic errors if the algorithm fails to generalize across different cases.
- Over-Reliance on Imaging Data Alone: Current AI models primarily rely on imaging data, excluding critical clinical information like patient history, ethnicity, lifestyle factors, and past treatments. This omission could lead to inaccuracies in real-world settings where a holistic view of the patient is necessary for accurate diagnosis.
- Lack of Real-World Testing: Many AI models have shown high accuracy in controlled settings but have not been rigorously tested in actual clinical environments. Applying these algorithms without adequate validation in real-world settings poses a risk of misdiagnosis.
- Dataset Limitations: Models trained on existing datasets may fail when exposed to skin conditions or lesion types not represented in the training data, potentially leading to errors when assessing atypical cases.
- Misdiagnosis Risk: The technology's limitations in handling unseen cases and the risk of overfitting to specific datasets increase the likelihood of false positives or negatives, especially if the AI is applied to skin lesions beyond its trained dataset scope.
- IFUs claims interpretation:
- Intended Purpose: The device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs, and interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: The device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus, and nails) and associated mucous membranes (conjunctival, oral, and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
- Contraindications: We advise not to use the device in: Skin structures located at a distance greater than 1 cm from the eye, beyond the optimal range for examination. Skin areas that are obscured from view, situated within skin folds or concealed in other manners, making them inaccessible for camera examination. Regions of the skin showcasing scars or fibrosis, indicative of past injuries or trauma. Skin structures exhibiting extensive damage, characterized by severe ulcerations or active bleeding. Skin structures contaminated with foreign substances, including but not limited to tattoos and creams. Skin structures situated at anatomically special sites, such as underneath the nails, requiring special attention. Portions of skin that are densely covered with hair, potentially obstructing the view and hindering examination.
- Operating Principle: The device is a computational medical tool leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures.
Article ID 35
- Title / Year: Diagnostic capacity of skin tumor artificial intelligence-assisted decision-making software in real-world clinical settings (2020)
- Author: Li, CX
- DOI / PMID: 10.1097/CM9.0000000000001002
- Brief Summary of the article: This study focuses on Youzhi AI, a software developed by Shanghai Maise Information Technology using deep learning algorithms to assist dermatologists in diagnosing skin tumors through image analysis. Trained on the extensive Chinese Skin Image Database, Youzhi AI employs a convolutional neural network architecture, incorporating a classification and segmentation system to identify skin tumor types. The software achieves a diagnostic accuracy of 91.2% for distinguishing between benign and malignant tumors, and 81.4% for identifying specific disease types, aligning with international standards in laboratory settings. In clinical trials, Youzhi AI's diagnostic accuracy was compared to that of dermatologists on clinical and dermoscopic images of skin tumors. While it showed no significant advantage over dermatologists in detecting malignancies (BMA), Youzhi AI surpassed dermatologists in disease type classification (DTA) for dermoscopic images. However, real-world performance variances are noted, as AI algorithms can yield lower diagnostic accuracy outside controlled environments. This analysis highlights Youzhi AI's potential to improve diagnostic consistency and support dermatologists, particularly in regions where diagnostic capabilities may be limited.
- Risks Interpretation:
- Diagnostic Variability: The article emphasizes that the diagnostic accuracy of AI systems like Youzhi AI may differ when tested in real-world conditions versus lab-controlled datasets. Performance metrics validated in research settings often drop when the AI encounters unfiltered clinical data, making diagnostic reliability a potential risk in diverse healthcare environments.
- Image Quality Sensitivity: AI systems may underperform if input images are of low quality, poorly focused, or obscured by factors like exogenous pigments. This sensitivity to image quality could lead to diagnostic inaccuracies if the images do not meet certain technical requirements.
- Interoperability with Different Databases: The software is primarily trained on the Chinese Skin Image Database (CSID), and its efficacy may be limited if applied to populations or conditions not adequately represented in the training dataset. Variability in performance across different populations and imaging technologies poses a risk of inconsistent diagnostic accuracy.
- Model Generalizability: The AI's dependency on a specific convolutional neural network (CNN) model (GoogLeNet Inception v4) trained under controlled conditions limits its adaptability to novel, untrained cases. Any updates or changes in clinical presentation may not be well-recognized by the software.
- Human-Software Interaction: The article highlights the need for dermatologists to be trained in specific image processing techniques (e.g., cropping) to optimize AI analysis. This dependency on correct human input introduces a risk of human error impacting diagnostic outcomes, especially if instructions are not followed precisely.
- IFUs claims interpretation:
- Intended Purpose: The device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs, and interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: The device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus, and nails) and associated mucous membranes (conjunctival, oral, and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
- Contraindications: We advise not to use the device in: Skin structures located at a distance greater than 1 cm from the eye, beyond the optimal range for examination. Skin areas that are obscured from view, situated within skin folds or concealed in other manners, making them inaccessible for camera examination. Regions of the skin showcasing scars or fibrosis, indicative of past injuries or trauma. Skin structures exhibiting extensive damage, characterized by severe ulcerations or active bleeding. Skin structures contaminated with foreign substances, including but not limited to tattoos and creams. Skin structures situated at anatomically special sites, such as underneath the nails, requiring special attention. Portions of skin that are densely covered with hair, potentially obstructing the view and hindering examination.
Article ID 38
- Title / Year: Automated detection of nonmelanoma skin cancer using digital images: a systematic review (2019)
- Author: Marka, A.
- DOI / PMID: 10.1186/s12880-019-0307-7
- Brief Summary of the article: The article explores the use of machine learning (ML) and artificial intelligence (AI) for diagnosing nonmelanoma skin cancer (NMSC) through image-based analysis, primarily using digital photography and dermatoscopy. The study addresses challenges in traditional visual diagnosis, where benign lesions often mimic malignancies, leading to invasive biopsies. Digital image-based ML models offer an alternative for early and accurate detection of NMSC, especially basal cell carcinoma (BCC) and cutaneous squamous cell carcinoma (CSCC), which are visually identifiable by skilled dermatologists. Various ML techniques are reviewed, including artificial neural networks (ANNs), decision trees, random forests, logistic regression, and support vector machines (SVMs). In some studies, models focus on identifying specific dermoscopic features of NMSC, such as telangiectasia, pink blush, and vascular characteristics, while others use global color and texture analysis across the entire lesion. The article highlights that most studies utilize separate training and test datasets or cross-validation to ensure accuracy, though some were criticized for overlapping data or using biased image sets. Accuracy metrics, such as the area under the receiver operating characteristic (AUROC) curve, were common, with AUROC values ranging from 0.832 to perfect classification. While some models reached 100% accuracy in specific studies, the results varied based on sample sizes and image sets. Overall, the review demonstrates the potential of AI and ML in dermatologic practice, though the quality of data sources and standardization of test methods remain areas for improvement.
- Risks Interpretation:
- Misclassification of Lesions: There is a risk of falsely classifying benign lesions as malignant, leading to unnecessary biopsies, treatments, and associated morbidity. This issue arises because benign lesions can mimic the appearance of NMSC, which can result in overtreatment.
- Generalization Errors: ML models may struggle to generalize effectively to novel cases not included in their training datasets. This could lead to inaccuracies in diagnosing skin conditions that differ from those represented in the training data.
- Insufficient Training Data: Some studies used small sample sizes or non-consecutive sampling methods, which may affect the robustness of the models. For instance, reliance on images from clinics or public datasets without ensuring diverse representation can limit the model's ability to perform well across varied populations.
- Quality of Input Data: The accuracy of the ML algorithms heavily relies on the quality of the input images. If images are poorly taken or not properly annotated, this can negatively impact the model's performance.
- Overfitting: Models that do not employ proper validation techniques may become too tailored to their training data (overfitting), resulting in poor performance on unseen data.
- Variability in Results: The article notes variability in reported metrics (e.g., sensitivity, specificity) across studies, which can lead to inconsistent outcomes and uncertainty about the reliability of the diagnostic tools.
- Ethical Considerations: The deployment of AI in clinical settings raises ethical concerns regarding accountability in case of diagnostic errors, as well as patient consent and data privacy issues.
- Dependence on Technology: Over-reliance on automated systems may diminish the role of experienced dermatologists, potentially leading to a decrease in the quality of clinical evaluations.
- IFUs claims interpretation:
- Intended Purpose: The device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs, and interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: The device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus, and nails) and associated mucous membranes (conjunctival, oral, and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
- Contraindications: We advise not to use the device in: Skin structures located at a distance greater than 1 cm from the eye, beyond the optimal range for examination. Skin areas that are obscured from view, situated within skin folds or concealed in other manners, making them inaccessible for camera examination. Regions of the skin showcasing scars or fibrosis, indicative of past injuries or trauma. Skin structures exhibiting extensive damage, characterized by severe ulcerations or active bleeding. Skin structures contaminated with foreign substances, including but not limited to tattoos and creams. Skin structures situated at anatomically special sites, such as underneath the nails, requiring special attention. Portions of skin that are densely covered with hair, potentially obstructing the view and hindering examination.
- Operating Principle: The device is a computational medical tool leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures.
- Use Environment: The device is intended to be used in the setting of healthcare organizations and their IT departments, which commonly are situated inside hospitals or other clinical facilities.
Article ID 58
- Title / Year: Melanoma and other skin lesion detection using smart handheld devices (2015)
- Author: Zouridakis, G.
- DOI / PMID: 10.1007/978-1-4939-2172-0_30
- Brief Summary of the article: Recent advancements in noninvasive diagnostic procedures for skin cancer have significantly transformed the landscape of dermatology, with dermoscopy emerging as the most widely adopted technique. This method enhances melanoma detection by more than 50% compared to traditional naked eye inspection, utilizing specialized equipment such as macro lenses for magnification and various light sources to visualize the skin's epidermal layers. Digital dermoscopy systems, equipped with automated image analysis capabilities, demonstrate impressive sensitivity and specificity rates for melanoma detection, often surpassing those of general practitioners and nearing the accuracy of dermatologists and dermoscopy experts. The development of portable dermoscopy devices has further expanded the accessibility of skin cancer screening. Devices like MelaFind and DermLite II utilize multispectral imaging techniques to capture multiple images of pigmented lesions and analyze their morphological characteristics. MelaFind, for example, can classify lesions based on their 3D disorganization in under a minute. The rise of smartphones has also played a pivotal role in delivering image-based diagnostic services, allowing users to capture and analyze skin lesions conveniently. Modern smartphones, with their advanced processors and high-resolution cameras, facilitate complex image analysis, enabling healthcare providers to deliver effective diagnostic services even in resource-limited settings. Smartphone applications for skin cancer screening have proliferated, providing tools for self-examination and lesion monitoring. Some applications incorporate algorithms to analyze images and predict the likelihood of malignancy. These apps often integrate with attachments designed for dermoscopic imaging, enabling users to obtain high-quality images for further analysis. The growing reliance on smartphones and digital technologies in dermatology reflects a broader trend toward personalized, accessible healthcare solutions. Automated classification systems have evolved significantly, employing objective dermoscopic criteria to evaluate pigmented lesions and aiding in early melanoma detection. Through image segmentation, feature extraction, and robust classification algorithms, these systems improve diagnostic accuracy, making them invaluable in contemporary dermatological practice.
- Risks Interpretation:
- Diagnostic Accuracy: While automated systems and smartphone applications can enhance diagnostic capabilities, there is a risk of misclassification or false negatives, potentially leading to missed melanoma cases or unnecessary biopsies.
- User Dependence: The effectiveness of smartphone applications relies heavily on the user's ability to capture high-quality images. Poor image quality or incorrect use of the technology may compromise diagnostic accuracy.
- Limited Training and Experience: General practitioners using these technologies may lack the necessary training or experience, which could affect their diagnostic skills compared to specialists. This can lead to variability in the interpretation of results.
- Data Privacy and Security: The use of smartphone applications raises concerns about data privacy and security, as personal health information may be at risk of exposure or misuse.
- Over-Reliance on Technology: There is a risk that clinicians might become overly reliant on automated systems and smartphone applications, potentially neglecting the importance of clinical examination and thorough patient history in the diagnostic process.
- Accessibility Issues: While portable devices and smartphone applications enhance accessibility, disparities in technology access and internet connectivity could create inequalities in skin cancer detection and treatment.
- IFUs claims interpretation:
- Intended Purpose: The device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing: quantification of intensity, count, extent of visible clinical signs, and interpretative distribution representation of possible International Classification of Diseases (ICD) categories.
- Indications: The device is indicated for use on images of visible skin structure abnormalities to support the assessment of all diseases of the skin incorporating conditions affecting the epidermis, its appendages (hair, hair follicle, sebaceous glands, apocrine sweat gland apparatus, eccrine sweat gland apparatus, and nails) and associated mucous membranes (conjunctival, oral, and genital), the dermis, the cutaneous vasculature and the subcutaneous tissue (subcutis).
- Contraindications: We advise not to use the device in: Skin structures located at a distance greater than 1 cm from the eye, beyond the optimal range for examination. Skin areas that are obscured from view, situated within skin folds or concealed in other manners, making them inaccessible for camera examination. Regions of the skin showcasing scars or fibrosis, indicative of past injuries or trauma. Skin structures exhibiting extensive damage, characterized by severe ulcerations or active bleeding. Skin structures contaminated with foreign substances, including but not limited to tattoos and creams. Skin structures situated at anatomically special sites, such as underneath the nails, requiring special attention. Portions of skin that are densely covered with hair, potentially obstructing the view and hindering examination.
Results from systematic review of literature
The clinical performance of artificial intelligence (AI) and software applications for assessing skin images, particularly in the context of skin cancer detection, has shown promising results across various studies. AI, specifically deep convolutional neural networks (CNNs), has been successfully implemented in primary care settings to enhance the sensitivity and specificity of melanoma detection. For instance, algorithms like Fotofinder Moleanalyzer have demonstrated sensitivity and specificity rates comparable to those of trained dermatologists (33). Additionally, a customized AI model developed to address data limitations achieved a high diagnostic accuracy and sensitivity in identifying melanoma, outperforming dermatologists in some instances (34).
Furthermore, studies have illustrated the potential of AI-driven algorithms to match or even surpass clinician performance in differentiating malignant from benign lesions, with large annotated datasets playing a critical role in training these models (35). The Youzhi AI software, trained on extensive databases, has shown diagnostic accuracies of 91.2% for distinguishing benign and malignant tumors, providing significant support in clinical settings (36). However, real-world performance can vary, and the reliance on high-quality datasets remains a challenge, particularly concerning class imbalances between melanoma and non-melanoma cases (37).
Recent advancements in smartphone applications and portable dermoscopy devices have also improved access to skin cancer screening, enabling patients to engage in self-examinations and facilitating remote diagnostic services (38). Although these technologies have made significant strides in accuracy and accessibility, challenges persist in ensuring consistent real-world implementation due to variations in user skill, patient demographics, and environmental factors. Overall, while AI and software applications present substantial opportunities for enhancing skin cancer diagnosis, further research and clinical integration are essential to maximize their efficacy and address existing limitations.
Post-market surveillance
We have prepared the post-market surveillance in compliance with Articles 83 to 86 and Annex III of MDR 2017/745.
Post-market surveillance plan
As part of its quality program AI Labs Group S.L. undertakes a systematic review of the performance of its products. The purpose of these reviews is to detect either infrequent complications or problems only apparent after widespread use or any long-term performance issues of devices in a post-market setting. The post-market surveillance process is based on information received from the field (e.g. customer complaint monitoring, feedback from sales representatives, reports from regulatory authorities, literature reviews, service information) and documented in the Quality System.
Pursuant to Article 83 of MDR, the PMS in place is a proactive and systematic process designed to monitor the performance of the medical device by collecting and analysing information relating to its use in the field. The PMS system is based on a PMS plan following requirement established in Annex III of MDR.
Serious incidents and FSCA
Serious incidents and field safety corrective actions (FSCAs) are documented, reported to regulatory authorities, investigated, and addressed according to specific procedures (GP-004
and GP-006
for incidents, GP-004
and SP-004-001
for FSCAs). Once a year, these are reviewed in-depth, and the PMS Report (R-TF-007-004
, for the class I legacy device) and PSUR (R-TF-007-003
, once the device is on the market) summarize incident counts, investigations, corrective actions, and FSCA statuses.
Non-Serious Incidents, Side Effects and Trend Analysis
Non-serious incidents are documented, evaluated and addressed according to the process described in GP-004
. Side effects not previously identified are reviewed yearly. Summaries of non-serious incidents, side effects, and improvement measures are included in the PMS Report and PSUR.
Non-serious incidents and anticipated side effects are analyzed for trends over time, following ISO/TR 20416 guidelines. If a trend is detected, a CAPA is initiated and the regulatory authorities notified accordingly. The results are documented in the PMS Report and PSUR.
Customer feedback
The post-market surveillance will evaluate product performance in cases through the collection of feedback from users. The primary objectives are to confirm product performance and positioning of the device. In addition, this is valuable to characterize and benchmark the performance against key competitive products and identify any unanticipated product performance issues.
According to the PMS Plan, feedback from users is documented, analysed and addressed and will be evaluated every year to identify any actions and improvements to be implemented.
Customer complaints
Customer complaints are reviewed for possible categorization as incidents. Serious incidents follow the reporting process described in GP-004
. Complaints are analyzed yearly and summarized in the PMS Report and PSUR.
Corrective and Preventive Actions
Documented under GP-006
, CAPAs are reviewed every three months to assess their status and effectiveness. An annual summary of CAPA outcomes and effectiveness is included in the PMS Report and PSUR.
Data on similar devices
Collected under GP-015
and Clinical Evaluation Plan, data on similar devices are reviewed annually, focusing on technical and clinical attributes, with additional information sourced from adverse events databases. Findings are documented in the Clinical Evaluation Report (CER) and PMCF evaluation report.
Regulatory requirements and clinical literature
Regulatory websites are reviewed annually for updated requirements that might affect clinical or risk assessments. Key findings are summarized in the PMS Report and PSUR.
Clinical literature is reviewed annually as part of PMCF activities per the Clinical Evaluation Plan. The findings are documented in the PMCF evaluation report and PMS Report.
Cybersecurity and State of the Art
Cybersecurity activities include monitoring new threats, incidents, and updates. ENISA and CISA are consulted for emerging cybersecurity threats. Relevant findings are documented in the PMS Report and PSUR.
Security Vulnerabilities of SOUPs and Software Tools
SOUPs (Software of Unknown Provenance) are monitored every six months, and security advisories for software tools are reviewed annually to identify vulnerabilities. Evaluation results are documented in respective DHF SOUP records and updated in PMS report and PSUR if necessary.
Sales data
Sales volume, usage frequency, user population characteristics, and country distribution are analyzed yearly. These data are included in the PMS Report and PSUR to help understand the device's impact and market performance.
Post Market Clinical Follow-up Plan
We are required to establish a process for Post-Market Clinical Follow Up to assess and confirm the product safety and performance of the device throughout its expected lifetime in accordance with the European Medical Device Regulation 2017/745. According to Annex XIV, Part B of this regulation, PMCF shall be understood to be a continuous process that updates the clinical evaluation. A PMCF plan and PMCF evaluation report or a justification why a PMCF is not applicable is required for all categories of CE marked devices. This section documents the determination of the need for a Post-Market Clinical Follow-Up (PMCF) plan or the justification.
As part of the PMS plan, a periodic safety update report (PSUR) will be performed on an annual basis, including a PMCF of the product considering the evidence related to the device that could have an impact on the quality, safety, or performance.
In accordance with MEDDEV 2.7/1 Revision 4, sufficient clinical evidence is documented in the prior clinical evaluation. In compliance to EU MDR, PMCF is required to collect direct clinical evidence to confirm continued performance and safety of the device in a clinical setting. A PMCF Plan (R-TF-007-002
) will be generated, and the results summarized in a PMCF Evaluation Report (R-TF-007-004
). The PMCF plan will establish the procedures and methods to proactively seek device indications for use and basic performance and safety data. The information collected and summarized in the PMCF report will be incorporated into the next clinical evaluation. The PMS will also check if the classification of the device is according with what's explained in this CER. In addition to PMCF, we have a robust quality system of capturing adverse events that identifies new risks and will seek to support the PMCF findings.
A PMCF Plan has been developed for the device to consistently collect relevant clinical data, confirming the benefit-risk determination in the Risk Management Report (R-TF-013-003
) and Clinical Evaluation Report (R-TF-015-003
). This includes assessing clinical safety, performance, any claims, and its alignment with the current state of the art.
PMCF Clinical Investigation
Following a proper premarket clinical evaluation, the decision to conduct specific PMCF studies must be based on the identification of possible residual risks and/or lack of clarity on long term clinical performance that may impact the benefit/risk ratio. PMCF specific activities may review issues such as long-term performance and/or safety, the occurrence of clinical events specific to defined patient populations, or the performance and/or safety of the device in a more representative population of users and patients.
The PMCF studies provide additional data to support the safety and effectiveness of the device in specific patient populations and under varying clinical conditions.
Pilot study for the clinical validation of an artifical intelligence algorithm to optimize the appropriateness of dermatology referrals
- Description:
- Design: Prospective observational and analytical study of a longitudinal clinical case series.
- Sample Size: This study pretends to include 400 patients.
- Duration: 4 months. An extension of the study has been requested to continue patient recruitment.
- Objective: To validate that Legit.Health artificial intelligence algorithms are a valid tool for optimizing the appropriateness of dermatology referrals.
- Additional information:
- Location: Health Center Sodupe-Güeñes, Health Center Balmaseda, Health Center Buruaga and Health Center Zurbarán, Spain.
- Current status: 79 patients enrolled as of september, 2024.
- Next steps: Continue patient enrollment.
- Rationale: In some cases, there are discrepancies between the diagnoses of primary care physicians and dermatologists ranging from 57% to 65% depending on the study. Data about how Legit.Health can improve the appropriateness of referrals to dermatology will be useful to assess the performance of Legit.Health in a real-world environment.
- Known limitations of the activity: The quantity and quality of the images collected, there is no control group and if the patient does not attend a visit with dermatologist.
- Search period: Study expeceted to finish Q2 2025
Evaluating the performance of Legit.Health in automated triage in teledermatology
- Description:
- Design: Observational and retrospective study.
- Sample size: 30000 images will be analyzed in this study.
- Objective: To assess the impact of implementing Legit.Health in reducing the average waiting time for skin cancer patients and to assess the sensitivity and specifity of Legit.Health detecting malignancy.
- Additional information:
- Location: Vall d'Hebron University Hospital, Spain.
- Rationale: Obtain performance data of Legit.Health by evaluating the clinical usefulness of instant imaging and the malignancy and urgency level in a population of 30000 referral images from primary care to dermatology.
- Known limitations: Quality of the images and subjectivity in diagnosis.
- Search period: 1 year.
Study for the clinical validation of a medical device for the priorization of consultations in patients with suspected skin cancer
- Description:
- Design: Prospective study with intervention.
- Sample size: 140 patients.
- Objective: To evaluate the impact of the medical device Legit.Health on prioritizing dermatology follow-up consultations in patients with suspected melanoma.
- Additional information:
- Location: Santa Creu i Sant Pau University Hospital, Spain.
- Rationale: Melanoma is a deadly disease and a late diagnosis reduces the chances of patient survivial. This study pretends to gather data of the performance of Legit.Health priorizing the control visits according to the suspicion of malignancy.
- Known limitations: Quantity and quality of the images, there is no control group, subjectivity in diagnosis and loss of follow-up.
- Search period: 1 year.
Study for the validation of a medical device for improving the diagnosis of skin conditions in Primary Care
- Description:
- Design: Prospective study with intervention.
- Sample size: Still to be confirmed.
- Objective: To validate the medical device Legit.Health to improve the diagnosis of skin conditions in Primary Care and confirmed by dermatology.
- Additional information:
- Location: Health Center Almozara, Health Center San Pablo, Health Center Revolvería and Miguel Servet University Hospital.
- Rationale: In some cases, there are discrepancies between the diagnoses of primary care physicians and dermatologists ranging from 57% to 65% depending on the study. Data about how Legit.Health can improve the agreement in the diagnoses between primary care and dermatology and improve the appropriateness of referrals.
- Known limitations: Quantity and quality of the images.
- Study period: 14 months.
Pilot study for the clinical validation of a medical device for the automatic assessment of severity and remote monitoring of patients with acne
- Description:
- Design: Prospective study with intervention.
- Sample size: 30 patients.
- Objective: To validate that the ALADIN severity scale developed by AI LABS GROUP S.L. measures the severity of facial acne with a capacity similar to or greater than that of a specialist, using a photograph taken with a smartphone.
- Additional information:
- Location: Dermatology DermoMedic Clinic, Spain.
- Rationale: Acne is one of the main reasons for dermatology consultations. This study aims to obtain performance data of Legit.Health by measuring the severity of facial acne.
- Known limitations: Quality of the images, the ALADIN scoring system focuses solely on lesion couting without making distinctions between the different types of lesions.
- Study period: 9 months.
Pilot study for the clinical validation of the SALT automatic system for measuring the severity of alopecia areata based on artificial intelligence
- Description:
- Design: Observational and prospective-retrospective study.
- Sample size: 30 patients.
- Objective: To validate that the automatic SALT severity measurement system for alopecia areata achieves an accuracy equal to or greater than that of an expert clinical using the "gold standard" SALT (Severity of Alopecia Tool).
- Additional information:
- Location: Still to be confirmed.
- Rationale: This study aims to collect data of Legit.Health performance filling up automatically the scoring system Severity of Alopecia Tool in a real-world environment.
- Known limitations: Quantity and quality of the images which can influence the precission.
- Study period: 3 months.
Pilot study for the clinical validation of an automatic EASI scoring system with artificial intelligence algorithms to assess the severity of atopic dermatitis
- Description:
- Design: Observational and retrospective study.
- Sample size: 100 images from different patients.
- Objective: To validate an automatic measurement system of the Eczema Area and Severity Index (EASI) based on artifical intelligence to determine the severity of atopic dermatitis, and that it does so with an accuracy simlar or better than the consensus of experts who use the "gold standard" EASI.
- Additional information:
- Location: Virgen de las Nieves University Hospital of Granada.
- Rationale: Atopic Dermatitis is one the most frequent conditions in dermatology. This study aims to collect data of Legit.Health performance filling up automatically the scoring system Eczema Area and Severity Index with images from patients of a real-world environment.
- Known limitations: Quality and quantity of imagens which can influence the precission.
- Study period: 3 months.
Pilot study for the clinical validation of a medical device for the quantification of severity and monitoring of the evolution of patients with FFA (Frontal Fibrosing Alopecia)
- Description:
- Design: Observational and prospective study.
- Sample size: 100 patients.
- Objective: To validate that the medical device Legit.Health is capable of measuring the severity of frontal fibrosing alopecia by automatically counting hairs and verify that it does so with a capacity equal or greater than the respective "gold standard" completed by the specialist.
- Additional information:
- Location: Ramón y Cajal University Hospital of Madrid, Spain.
- Rationale: Measuring the severity of frontal fibrosing alopecia is a subjective process that does not allow for the detection of small changes. This study aims to collect performance data of Legit.Health in measuring the severity of Frontal Fibrosing Alopecia.
- Known limitations: Quality and quantity of images which can influence the precision.
- Study period: 2 years.
Pilot study for the clinical validation of a medical device for the authomatic triage in teledermatology
- Description:
- Design: Prospective interventional study.
- Sample size: Still to be confirmed.
- Objective: To validate that the medical device Legit.Health is capable to priorize the referrals from primary care to dermatology according to the severity.
- Additional information:
- Location: Vall d'Hebron University Hospital, Spain.
- Rationale: This study aims to collect data about how Legit.Health can improve the referrals from primary care to dermatology according to the suspiction of malignancy.
- Known limitations: The quality and quantitu of images, which can affect the precision, there is no control group, the patient does not attend a visit with dermatologist.
- Study period: 14 months.
Compliance with applicable regulatory requirements
As previously indicated, one of the main objectives of this clinical evaluation is the demonstration of compliance with the applicable GSPRs as indicated in MDR. The specific GSPRs to be supported with clinical evidence have been identified within the CEP (R-TF-015-001 Clinical Evaluation Plan
).
Requirement of safety (GSPR 1)
As defined in the CEP, the following parameters need to be addressed to determine the safety of the device:
- Patient Safety: Minimizing risks of misdiagnosis and misinterpretation of imaging data to ensure accurate clinical information for healthcare practitioners.
- Data Privacy and Security: Implementing state-of-the-art security measures to protect sensitive patient information, ensuring compliance with data protection regulations.
- Prevention of Over-Reliance on Imaging Data: Encouraging informed decision-making by healthcare practitioners and emphasizing the importance of integrating imaging data with clinical expertise.
- Variability in Skin Condition Presentations: Accommodating intra- and inter-class variability to ensure accurate assessment of a wide range of skin conditions.
- Diversity of Dataset: Addressing limitations in dataset diversity to enhance the algorithm's generalizability and applicability across different patient populations.
- Robust Verification and Validation: Conducting thorough verification and validation processes, including real-world testing, to confirm consistent performance in various clinical environments.
The Risk Management process is conducted in accordance with principles of ISO 14971:2019 Application of Risk Management to Medical Devices. The device has been reviewed for risk and undergone a failure modes and effects analysis (FMEA) that is described in a general risk assessment report. In preparing the risk assessment, the factors considered include data privacy and security, potential over-reliance on imaging data by healthcare practitioners, intra- and inter-class variability in skin condition presentations, limitations in dataset diversity, risks of misinterpretation of imaging data, lack of real-world testing, and the overall accuracy and reliability of the device's algorithm in diverse clinical scenarios.
The previous generation of the device became available by the end of 2020, with 21 contracts signed, covering both remote and on-site usage in government-run and for-profit healthcare providers. Over 4500 reports have been generated by more than 500 practitioners, assisting over 1000 patients during this period. Importantly, no serious incidents or Field Safety Corrective Actions (FSCA) were reported for the device during the reviewed period. Additionally, no significant trends or deviations were observed in the device's safety and performance, as the focus remained on its previous generation. The Post-Market Clinical Follow-Up (PMCF) evaluation reported consistent positive user feedback, with both primary and secondary healthcare professionals expressing satisfaction with the device.
The Post-Market Clinical Follow-Up (PMCF) activities highlighted key findings, including a comprehensive review of clinical literature, which validated the safety of the device in accordance with current scientific knowledge. PMCF studies further confirmed the device's safety, performance, and clinical benefits in real-world settings. No new risks or adverse events were identified, supporting the ongoing use of the device. Success metrics from the device's image recognition processor validated its operational efficacy, while user feedback confirmed the device's functionality as intended. A comparison with similar devices on the market and consultations with sanitary alert databases further supported its safety profile and risk management efforts. Additionally, a detailed analysis of complaints showed that corrective and preventive actions (CAPAs) have successfully addressed concerns, contributing to the device's improvement. PMCF findings will continue to inform clinical evaluation and risk management, ensuring the device's sustained safety and efficacy.
Regarding corrective and preventive actions (CAPAs), 30 were opened in the period 2020 - 2023. Of these, 22 CAPA are related to non-conformity in our QMS processes, 8 CAPA are related to non-conformity in our device. The most significant non-conformity, R-006-001-56, was due to a performance issue with the AI analysis of images. On May 2, 2023, Consultant Connect reported discrepancies in results when analyzing 50 photos, with only 8 images providing accurate diagnoses. The investigation revealed that the issue stemmed from poor-quality images that did not meet the minimum standards for accurate AI analysis. The AI was ultimately determined to be functioning correctly, and the customer was educated on the proper image quality and centering techniques. Additionally, minor observations were made regarding ICD-coded diseases not being properly documented in the Device History File (DHF), and FDA audit findings were recorded incorrectly in the CAPA tool. A recommendation for a training matrix was also noted during the audit. Despite these issues, no serious safety concerns were raised, and the CAPAs implemented have led to improvements in the device's performance.
Risks identified in the Literature Search Report and Vigilance Databases
As per MDR 2017/745 recommendations, currently known safety outcomes were also contemplated in the Literature Search. Keywords identified with regards to safety and performance were the following: ("performanc*" OR "safe*" OR "clinical").
The clinical search and vigilance database review for devices similar to the device, identified several key risks relevant to the use of AI in dermatological assessments. Data limitations and potential biases due to non-representative datasets were noted, which may lead to reduced accuracy for underrepresented skin types and conditions. False positives and negatives were common risks, with AI sometimes misclassifying benign lesions as malignant or vice versa, potentially leading to unnecessary treatments or missed diagnoses. Dependence on image quality is a critical factor, as poor-quality images can compromise AI performance, especially in remote or teledermatology settings.
Additional risks include over-reliance on AI by non-specialists, which may lead to reduced diagnostic vigilance, and challenges with generalizing AI performance from controlled settings to real-world environments, where performance can vary significantly. Legal and ethical concerns, such as liability in the event of diagnostic errors, were also raised, alongside issues related to data privacy and security. Collectively, these risks highlight the importance of using diverse, high-quality datasets, thorough validation, and maintaining human oversight to support AI-driven dermatological tools effectively and safely in clinical practice.
Importantly, no adverse events or recalls were identified for the similar devices (MoleScope and SkinVision) across FDA, Swissmedic, MHRA, and AEMPS databases over the last decade, indicating an acceptable safety profile despite the noted risks.
Residual risks identified
After hazard analysis, through a Severity Rating Number, hazards were assigned either Acceptable
, As far as possible
, or Not acceptable
.
The residual risks identified in the risk analysis encompass several key concerns, each mitigated through specific actions and assessed for severity and likelihood. For incorrect diagnoses or follow-ups, mitigations include defining intended users (healthcare professionals) in the IFU and enhancing AI model supervision through explainability metrics. This risk has a severity of 3 and a likelihood of 2, leading to an As far as possible
residual risk status.
Risks related to image quality such as artefacts, resolution, and inadequate lighting conditions are addressed by a pre-processing algorithm that validates image quality, provides scoring, and prompts users to retake images if necessary. User training is also provided. These risks have a severity of 3 and likelihood of 2, indicating residual risks are managed As far as possible
.
For potential data transmission, input, and accessibility failures, the device employs state-of-the-art security and software availability techniques to ensure data integrity and accessibility. These data-related risks are similarly evaluated with a severity of 3 and likelihood of 2, leading to a residual risk level of As far as possible
across all categories.
Requirement of performance (GSPR 1)
As defined in the CEP, the following parameters need to be addressed to determine the performance:
- Diagnostic Accuracy: Evaluate how accurately the device identifies dermatological conditions compared to clinical diagnoses.
- Clinical Utility: Assess the device's effectiveness in improving patient management and outcomes, such as reducing the need for in-person consultations.
- User Satisfaction: Measure the satisfaction levels of both healthcare providers and patients regarding the device's usability and effectiveness.
- Referral Efficiency: Analyze the impact of the device on the rate of unnecessary referrals to dermatology specialists.
- Workflow Integration: Determine how well the device integrates into existing clinical workflows and its effect on overall efficiency.
- Sensitivity and Specificity: Assess the device's ability to correctly identify true positive and true negative cases for various skin conditions.
The ability to achieve its intended purpose as stated by the manufacturer has been demonstrated through this CER.
The device has established, implemented, documented and maintained a risk management system in accordance with Annex I of the MDR. This involves the identification and assessment of risks, the implementation of risk control measures and the verification of the acceptability of residual risks.
We have performed a risk analysis of the product and has concluded that the overall residual risk is acceptable after all risk control measures have been implemented and verified. Updated warnings have been established in the Instructions for Use (IFU), providing revised safety information in accordance with current legislation.
The device effectively meets several critical performance parameters essential for dermatological assessments in professional healthcare environments. The clinical evaluation demonstrated that the device exhibits high diagnostic accuracy, correctly identifying a range of skin conditions with sensitivity and specificity rates exceeding established benchmarks. Additionally, the device has been shown to enhance clinical utility by streamlining the assessment process, thereby reducing the need for unnecessary referrals to specialists. User satisfaction surveys indicated a positive response from both healthcare providers and patients, highlighting the device's usability and effectiveness in improving patient management. Furthermore, the integration into existing workflows has resulted in improved efficiency, allowing clinicians to focus more on patient care. Overall, the device not only supports accurate diagnostics but also enhances the overall patient experience in dermatological care.
In terms of post-market clinical investigations, the completed and ongoing clinical studies provide valuable data regarding the device's performance. One of the key clinical studies, the clinical validation study of a CAD system for melanoma detection, achieved an AUC of 0.842 in identifying melanoma and 0.8983 in detecting malignancy, with a sensitivity of 90.32% and specificity of 80.54%. Another study assessing the clinical utility of the device in continuous and remote monitoring of patient condition severity yielded a mean score of 76.67 on the Clinical Utility Questionnaire (CUS), reflecting positive feedback despite a small sample size.
The optimization study of clinical flow in dermatology, performed in 2023, demonstrated improvements in diagnostic accuracy, with the device achieving a sensitivity of 81% and specificity of 52% for detecting malignancy. The prospective analysis, set to continue in 2025, has shown further improvements, with an AUC of 0.94, sensitivity of 88%, and specificity of 85%. Furthermore, the device has been shown to increase diagnostic accuracy in primary care by 27.02% and by 10.46% in dermatology, with a reduction in referrals to dermatology by over 50%.
Additional studies, such as the non-invasive pilot study on generalized pustular psoriasis, have demonstrated increases in diagnostic accuracy for both dermatologists and primary care physicians. The device has also proven useful in reducing unnecessary referrals, supporting its integration into both primary care and dermatology settings.
In conclusion, the PMS activities and the clinical investigations support the continued safe and effective use of the device. The CAPAs, feedback analysis, and clinical study results highlight ongoing improvements, ensuring that the device remains compliant with GSPR 1 and continues to provide clinical value.
Requirement of acceptable benefit/risk ratio (GSPR 2, 6)
As indicated in the CEP, the benefit/risk ratio can be considered acceptable if:
- All applicable general safety requirements are met based on the pre-established safety parameters.
- All performance related GSPRs are met based on the pre-set performance parameters.
- Any residual risk with clinical impact is identified in the risk assessment.
Bench testing, pre-clinical testing, and manufacturing activities have been performed to reduce risks of the device. For the identified hazards, it is considered that the actions performed to control and reduce the hazards, result in residual risks that are acceptable when weighed against the product benefits. Further, the actions taken to reduce the risks identified do not pose any additional risks in themselves. Outcomes of the risk-benefit analysis have shown that the potential patient benefits outweigh potential patient risks, as evidenced in the summary of adverse and beneficial events. A risk factor was respectively calculated for each adverse and beneficial event and summed. The total risk score allows for a conclusion to be made that the benefit outweighs the risks.
Other manufacturers have placed similar devices in the market for many years without reporting any critical incidences.
Because of this clinical evaluation, the conclusions are that the device is indicated to support for health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by:
- Providing quantifiable data on the intensity, count and extent of clinical signs such as erythema, desquamation, and induration, among others.
- Providing an interpretative distribution representation of possible International Classification of Diseases (ICD) categories that might be represented in the pixels content of the image.
According to the MDR 2017/745 legislation, the benefit-risk is assessed comprehensively and both safety and performance parameters are considered. The device is an effective and safe product aimed to be used to give support for health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing quantifiable data on the intensity, count and extent of clinical signs such as erythema, desquamation, and induration, among others, and an interpretative distribution representation of possible International Classification of Diseases (ICD) categories that might be represented in the pixels content of the image. The clinical evidence demonstrates conformity with the relevant GSPRs.
The performance and safety of the device, as claimed, have been established when the product is used under the conditions and for the purposes intended, and the risks associated with the use of the device are acceptable when weighed against the benefits to the patient.
The data from the literature assessed demonstrates that The device offers a balanced risk/benefit profile for dermatological assessments. Clinical studies indicate that the device significantly enhances diagnostic accuracy, with validation results showing high sensitivity and specificity in detecting various skin conditions. The literature reveals that while there are potential risks related to data quality and variability, these are mitigated by the device's robust algorithms and ongoing performance monitoring. Furthermore, the evidence suggests that the benefits, such as improved patient management and timely interventions, substantially outweigh these risks. Overall, the findings support the conclusion that The device provides meaningful clinical value while maintaining an acceptable safety profile.
According to the period of this report, the literature data assessed has not changed the risk characterisation or understanding of the medical device product. Although there are risks associated with the use of the device, there is an occasional, remote, or low probability for their occurrence, and thus the residual risk evaluation is considered acceptable. The IFU contains sufficient information to reduce and mitigate any possible risks.
All things considered, the three requirements (mentioned above) for an acceptable benefit/risk ratio are met, and it is concluded that the clinical data on benefits outweigh the risks when the device is used to give support for health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing quantifiable data on the intensity, count and extent of clinical signs such as erythema, desquamation, and induration, among others, and an interpretative distribution representation of possible International Classification of Diseases (ICD) categories that might be represented in the pixels content of the image. Therefore, the benefit/risk balance of this product is considered positive.
Requirement on acceptability of undesirable side-effects (GSPR 8)
A review of the clinical data shows that no new undesirable side-effects or emerging risks have been identified through this clinical evaluation. The reviewed clinical data does not suggest any further risk mitigation or required amendments to the product information, IFU, or warnings. Regarding the patient harms associated with risk hazards as noted in the risk documentation, the rates of occurrence found within the clinical data on the device in this CER were within the anticipated ranges in the benchmark occurrence in the published literature.
Requirement on devices that incorporate software or for software that are devices in themselves (GSPR 17.2)
The compliance with the General Safety and Performance Requirements (GSPR) regarding software development and manufacturing can be justified through several key aspects. First, the software incorporates the latest advancements in computer vision algorithms and image processing techniques, ensuring that it remains at the forefront of technological innovation in dermatological assessments.
Additionally, the software development process follows a structured life cycle that includes essential phases such as requirements analysis, design, implementation, testing, and deployment. This systematic approach guarantees thorough documentation and traceability throughout the software's development, promoting accountability and quality assurance.
A comprehensive risk management process, aligned with ISO 14971, has been established to identify, evaluate, and mitigate risks associated with the software's use. This process addresses potential risks related to incorrect diagnoses, data security, and user interactions, ensuring that all identified risks are appropriately managed.
To further enhance compliance, the software employs robust security measures to protect sensitive patient data during transmission and storage. This includes the use of encryption protocols, secure data access controls, and adherence to industry standards for data protection, all of which safeguard against unauthorized access and ensure data integrity.
Finally, a rigorous verification and validation strategy is in place to ensure that the software meets specified requirements and performs as intended. This includes comprehensive unit testing, integration testing, and clinical validation with real-world data, confirming the accuracy and reliability of the software's outputs in clinical practice. By adhering to these principles, the device demonstrates compliance with the GSPR, ensuring it is developed and manufactured in a manner that prioritizes safety, performance, and reliability in its intended use.
Answer to specific performance and safety questions
The clinical questions have been answered as follow after the analysis of all the clinical performance data available.
1. Does the device effectively assist healthcare professionals in evaluating dermatological conditions by analyzing images of the skin structures, providing data that supports clinical decision-making?
The device effectively assists healthcare professionals in evaluating dermatological conditions by analyzing images of skin structures. The clinical evaluation demonstrates that the device enhances diagnostic accuracy, supporting clinical decision-making through detailed data analysis.
Clinical studies reveal that the device achieves high sensitivity and specificity in identifying various skin conditions, providing healthcare practitioners with valuable insights to aid in patient assessments. The device processes images to generate clinical data, which helps practitioners evaluate skin conditions more thoroughly and make informed decisions regarding patient management.
Moreover, the literature indicates that the device's performance is complemented by robust algorithms that account for intra- and inter-class variability, thereby addressing potential risks associated with data quality. Overall, the evidence supports the conclusion that the device plays a crucial role in dermatological assessments, facilitating improved patient care and outcomes.
2. Does the device consistently support the monitoring of skin conditions over time, enabling accurate longitudinal assessment for healthcare providers?
The device consistently supports the monitoring of skin conditions over time, enabling accurate longitudinal assessment for healthcare providers. The clinical evaluation highlights that the device's image analysis capabilities allow for ongoing tracking of dermatological changes, facilitating comprehensive assessments of patients' skin conditions across multiple visits.
Studies demonstrate that the device effectively captures and analyzes sequential images, providing data that helps healthcare professionals evaluate the progression or improvement of skin conditions over time. This longitudinal monitoring is essential for making informed clinical decisions regarding treatment efficacy and patient management.
Additionally, the device's algorithms are designed to account for variability in skin appearances, ensuring reliable comparisons between assessments at different time points. The accumulated evidence confirms that the device enhances the ability of healthcare providers to perform consistent and accurate monitoring of skin conditions, ultimately contributing to better patient outcomes.
3. Does the device maintain the quality and accuracy of its image analysis and data processing to ensure reliable support for clinical evaluations?
The device maintains high quality and accuracy in image analysis and data processing, supporting reliable clinical evaluations. Clinical studies conducted on the device confirm its capability to detect and analyze skin conditions by accurately processing images to highlight key clinical signs, such as erythema, edema, and lesion counts. According to the evaluations, the device demonstrated consistency in processing images with minimal variability, even in diverse skin types and lighting conditions. Furthermore, comparisons with similar devices, including MoleScope and SkinVision, underscore that the device meets or exceeds typical accuracy benchmarks for image-based dermatological tools, as noted in recent SOTA reviews for these devices
4. Does the device exhibit an acceptable safety profile, with a risk level comparable to or lower than other AI-based dermatological assessment tools commonly used in clinical practice?
The device has an acceptable safety profile, with risk levels comparable to or lower than similar AI-based dermatological assessment tools. Its safety assessment primarily focuses on minimizing risks related to potential misdiagnosis, image quality variability, and data security, as outlined in the IFU. Studies confirm that the device mitigates these risks effectively by incorporating image quality control features and extensive algorithm training on diverse datasets, reducing the likelihood of errors. Its risk profile aligns closely with, or improves upon, that of other similar tools, which is widely accepted in clinical practice.
5. Does the device demonstrate equivalent or superior performance compared to other dermatological analysis tools in providing reliable clinical data to support patient evaluations?**
The device demonstrates equivalent or, in some cases, superior performance compared to other dermatological analysis tools. Clinical validation studies show that the device excels in recognizing and quantifying detailed clinical signs relevant to skin conditions, such as the specific characteristics of erythema, crusting, and edema, among others. This level of detail not only supports healthcare providers with more comprehensive data for patient evaluations but also positions the device as a more robust option for tracking conditions longitudinally. Comparative analysis with other tools indicates that the device offers an enhanced ability to process complex images, leading to a more reliable dataset for clinical decision-making.
Conclusions
We have performed a clinical evaluation to support approvals and registrations. There is sufficient data available from its clinical use to demonstrate safety and performance. Specific regulations, standards, guidance documents, and recommendations are followed in the preparation of this report.
Compliance with the applicable general safety and performance requirements (GSPRs) is mainly demonstrated based on three key components: valid clinical association, technical performance, and clinical performance, as outlined in the Clinical Evaluation Plan (CEP). The Clinical Evaluation Report (CER) for the device includes evidence from safety and performance data sourced from non-clinical testing, such as pre-clinical and bench testing results that verify the product's design and functional integrity. Additionally, the CER presents an in-depth analysis of the valid clinical association, examining the relationship between visible skin structure abnormalities and the International Classification of Diseases (ICD) categories through a thorough literature review, screening, and data appraisal. Clinical performance is also assessed, covering identified relevant data, risk management, and evaluation of clinical safety and performance through both manufacturer-sourced and literature-based evidence. This structured assessment confirms that the device meets the GSPRs by ensuring safety, effective performance, a favorable benefit/risk ratio, and the reduction of undesirable side effects.
Preclinical testing confirmed that the device performs well throughout various critical aspects, including the technical verification and validation of design requirements, cybersecurity protections, and usability engineering. The preclinical bench testing validated that the device's machine learning models, spanning image recognition, object detection, and semantic segmentation—met their respective performance metrics for sensitivity, accuracy, and precision, essential for reliable skin condition assessment. Usability engineering evaluations demonstrated that the device integrates seamlessly within healthcare IT systems via an API, delivering efficient data handling and robust user interaction through compliance with healthcare interoperability standards. Cybersecurity assessments addressed potential risks associated with AI model vulnerabilities and external access threats, ensuring that security by design principles are incorporated to safeguard patient data integrity and device functionality. Collectively, these assessments confirm that the device meets technical performance criteria, fulfilling its intended purpose with validated safety, usability, and security across operational conditions.
Comprehensive literature searches were conducted in accordance with a systematic literature review methodology. For the state of the art, findings highlighted the role of AI and machine learning in advancing dermatological diagnostics for conditions like melanoma, basal cell carcinoma, and acne vulgaris. Melanoma remains a critical focus due to its aggressive nature, and early detection is crucial for improving patient outcomes. Various AI-based tools demonstrated high sensitivity and accuracy in distinguishing melanoma from benign lesions, with models like AcneGrader addressing acne severity. WHO's ICD-11 Classification of Dermatological Diseases provides an enhanced framework for skin disease categorization, facilitating better patient record integration and data analysis. Studies also identified multiple AI-driven approaches for skin condition detection, including CNN-based frameworks, which improve diagnostic accuracy and mitigate limitations of traditional methods. However, the use of AI in dermatology poses risks like diagnostic inaccuracy, data bias, and overreliance on technology, underscoring the need for algorithm refinement, comprehensive datasets, and clinician training.
In clinical performance, AI applications have shown promising sensitivity and specificity for skin cancer detection, sometimes matching or exceeding dermatologist performance in distinguishing benign from malignant lesions. The use of annotated datasets was emphasized for optimizing model training, though challenges remain regarding dataset quality and class imbalance. Advances in smartphone applications and portable devices have also broadened access to skin screenings, supporting remote diagnostics. Despite these advancements, real-world effectiveness may vary due to patient demographics and environmental factors, highlighting the need for continued research to ensure AI's reliability and integration into clinical practice.
The remaining risks that have been identified associated with diagnostic accuracy, image quality, and data security cannot be mitigated further and are considered acceptable when weighed against the benefits to the patient. All harms have been defined with their potential causes of failure and associated mitigation activities.
The IFU clearly explains the intended use of the device, including its indications, contraindications, user instructions, and essential safety information to ensure proper operation and effective patient management.
The Post-Market Surveillance (PMS) report for the legacy device confirms that the device's safety and performance remain stable, with no serious incidents or Field Safety Corrective Actions (FSCA) reported since its launch in 2020. User feedback remains positive, and Post-Market Clinical Follow-Up (PMCF) studies have validated its safety and efficacy, confirming its clinical benefits in real-world settings. Corrective and Preventive Actions (CAPAs) were addressed, notably improving image analysis performance by enhancing image quality standards. Clinical investigations show strong diagnostic accuracy, with the device increasing diagnostic precision in dermatology and reducing unnecessary referrals. Overall, the device continues to meet regulatory requirements and supports effective healthcare practices.
In conclusion, it has been shown that there is sufficient evidence to establish the safety and performance of the device when used in accordance with the IFU. The data are adequate to assess the benefits and risks associated with the subject device, concluding that the benefit-risk profile is acceptable. Therefore, this initial clinical evaluation demonstrates that the available clinical data are sufficient to establish conformity with all applicable General Safety and Performance Requirements (Annex I) of the Regulation (EU) 2017/745 of the European Parliament and of the Council of 5 April 2017 on medical devices (MDR) and confirm the safety and performance of the device. With this Clinical Evaluation, the device has been confirmed to be within the current state-of-the-art practice.
Date of next clinical evaluation
According to section 6.2.3 of MEDDEV 2.7/1 revision 4 (June 2016) in the Guidelines on Medical Devices, the clinical evaluation is actively updated:
- when the manufacturer receives new information from PMS that has the potential to change the current evaluation;
If no such information is received, then
- at least annually if the device carries significant risks or is not yet well established; or
- every 2 to 5 years; if the device is not expected to carry significant risks and is well established, a justification should be provided.
According to this information, it is proposed to update the clinical evaluation report annually. Therefore, the next CER for the device will be due in 2025.
Bibliography
- Papachristou P, Söderholm M, Pallon J, Taloyan M, Polesie S, Paoli J, et al. Evaluation of an artificial intelligence-based decision support for the detection of cutaneous melanoma in primary care: a prospective real-life clinical trial. British Journal of Dermatology. 2024 Jul 1;191(1):125-33.
- Shafiq M, Aggarwal K, Jayachandran J, Srinivasan G, Boddu R, Alemayehu A. A novel Skin lesion prediction and classification technique: ViT-GradCAM. Skin Research and Technology. 2024 Sep 1;30(9).
- Tajerian A, Kazemian M, Tajerian M, Malayeri AA. Design and validation of a new machine-learning-based diagnostic tool for the differentiation of dermatoscopic skin cancer images. PLoS One. 2023 Apr 1;18(4 April).
- Sangers T, Reeder S, Van Der Vet S, Jhingoer S, Mooyaart A, Siegel DM, et al. Validation of a Market-Approved Artificial Intelligence Mobile Health App for Skin Cancer Screening: A Prospective Multicenter Diagnostic Accuracy Study. Dermatology. 2022 Jul 1;238(4):649-56.
- Liu S, Fan Y, Duan M, Wang Y, Su G, Ren Y, et al. AcneGrader: An ensemble pruning of the deep learning base models to grade acne. Skin Research and Technology. 2022 Sep 1;28(5):677-88.
- Xia M, Kheterpal MK, Wong SC, Park C, Ratliff W, Carin L, et al. Lesion identification and malignancy prediction from clinical dermatological images. Sci Rep. 2022 Dec 1;12(1).
- White JML, Lui H, Chute CG, Jakob R, Chalmers RJG. The WHO ICD-11 Classification of Dermatological Diseases: a new comprehensive online skin disease taxonomy designed by and for dermatologists. Br J Dermatol [Internet]. 2022 Jan 1 [cited 2024 Oct 31];186(1):178-9. Available from: https://pubmed.ncbi.nlm.nih.gov/34289080/
- Giavina-Bianchi M, de Sousa RM, de Almeida Paciello VZ, Vitor WG, Okita AL, Prôa R, et al. Implementation of artificial intelligence algorithms for melanoma screening in a primary care setting. PLoS One. 2021 Sep 1;16(9 September).
- Muñoz-López C, Ramírez-Cornejo C, Marchetti MA, Han SS, Del Barrio-Díaz P, Jaque A, et al. Performance of a deep neural network in teledermatology: a single-centre prospective diagnostic study. Journal of the European Academy of Dermatology and Venereology. 2021 Feb 1;35(2):546-53.
- Raj R, Londhe ND, Sonawane R. Automated psoriasis lesion segmentation from unconstrained environment using residual U-Net with transfer learning. Comput Methods Programs Biomed. 2021 Jul 1;206.
- Maqsood S, Damaševičius R. Multiclass skin lesion localization and classification using deep learning based features fusion and selection framework for smart healthcare. Neural Networks. 2023 Mar 1;160:238-58.
- Thomsen K, Iversen L, Titlestad TL, Winther O. Systematic review of machine learning for diagnosis and prognosis in dermatology. Journal of Dermatological Treatment. 2020 Jul 3;31(5):496-510.
- Li Y, Shen L. Skin lesion analysis towards melanoma detection using deep learning network. Sensors (Switzerland). 2018 Feb 11;18(2).
- Manuskiatti W, Wongdama S, Viriyaskultorn N, Li JB, Kulthanan K, Techapichetvanich T. Long-term efficacy and safety of nonablative monopolar radiofrequency in the treatment of moderate to severe acne vulgaris. Lasers Surg Med. 2024 Feb 1;56(2):133-41.
- Nakamura M, Asaka Y, Ogawara T, Yorozu Y. Nipple Skin Trauma in Breastfeeding Women during Postpartum Week One. Breastfeeding Medicine. 2018 Sep 1;13(7):479-84.
- Oyama S, Funasaka Y, Tsuchiya SI, Kawana S, Saeki H. Increased number of mast cells in the dermis in actinic keratosis lesions effectively treated with imiquimod. Journal of Dermatology. 2017 Aug 1;44(8):944-9.
- Kim BR, Won SH, Kim JW, Kim M, Jeong J Il, Shin JW, et al. Development of a new classification and scoring system for scalp conditions: Scalp Photographic Index (SPI). Journal of Dermatological Treatment. 2023;34(1).
- Bayramova A, Mane T, Hopkins C, Zheng Y, Ogunleye TA, Taylor SC, et al. Photographing Alopecia: How Many Pixels Are Needed for Clinical Evaluation? J Digit Imaging. 2020 Dec 1;33(6):1404-9.
- Gudobba C, Mane T, Bayramova A, Rodriguez N, Castelo-Soccio L, Ogunleye TA, et al. Automating Hair Loss Labels for Universally Scoring Alopecia from Images: Rethinking Alopecia Scores. JAMA Dermatol. 2023 Feb 15;159(2):143-50.
- Bruni F, Alessandrini A, Starace M, Orlando G, Piraccini BM. Clinical and trichoscopic features in various forms of scalp psoriasis. Journal of the European Academy of Dermatology and Venereology. 2021 Sep 1;35(9):1830-7.
- Yafi A, Muakkassa FK, Pasupneti T, Fulton J, Cuccia DJ, Mazhar A, et al. Quantitative skin assessment using spatial frequency domain imaging (SFDI) in patients with or at high risk for pressure ulcers. Lasers Surg Med. 2017 Nov 1;49(9):827-34.
- Kim IS, Kim BR, Youn SW. Differentiation of Jessner's Lymphocytic Infiltration of the Skin from Various Chronic Cutaneous Lupus Erythematosus Subtypes by Quantitative Computer-Aided Image Analysis. Dermatology. 2016 Feb 1;232(1):57-63.
- Tao M, Li M, Zhang Y, Liu Y, Jiang P, Liu Y, et al. Objectively quantifying facial erythema in rosacea aided by the ImageJ analysis of VISIA red images. Skin Research and Technology. 2023 Jan 1;29(1).
- Logger JGM, de Jong EMGJ, Driessen RJB, van Erp PEJ. Evaluation of a simple image-based tool to quantify facial erythema in rosacea during treatment. Skin Research and Technology. 2020 Nov 1;26(6):804-12.
- Pakdaman SF, Samadi A, Fattahi M, Naeimifar A, Ardehali FA, Ketabi Y, et al. Fabrication and efficacy assessment of combination of brimonidine and ivermectin for treatment of papulopustular rosacea. J Cosmet Dermatol. 2024 Sep 1;23(9):2973-81.
- Frew J, Penzi L, Suarez-Farinas M, Garcet S, Brunner PM, Czarnowicki T, et al. The erythema Q-score, an imaging biomarker for redness in skin inflammation. Exp Dermatol. 2021 Mar 1;30(3):377-83.
- Foolad N, Prakash N, Shi VY, Kamangar F, Wang Q, Li CS, et al. The use of facial modeling and analysis to objectively quantify facial redness. J Cosmet Dermatol. 2016 Mar 1;43-8.
- Schaller M, Riel S, Bashur R, Kurup N, Schnidar H, Fehrenbacher B. Ivermectin treatment in rosacea: How novel smartphone technology can support monitoring rosacea-associated signs and symptoms. Dermatol Ther. 2022 Nov 1;35(11).
- Montero-Vilchez T, Diaz-Calvillo P, Rodriguez-Pozo JA, Cuenca-Barrales C, Martinez-Lopez A, Arias-Santiago S, et al. The burden of hidradenitis suppurativa signs and symptoms in quality of life: Systematic review and meta-analysis. Int J Environ Res Public Health. 2021 Jul 1;18(13).
- Wu MY, Wang CH, Ng CY, Kuo TT, Chang YC, Yang CH, et al. Periorbital erythema and swelling as a presenting sign of lupus erythematosus in tertiary referral centers and literature review. Lupus. 2018 Oct 1;27(11):1828-37.
- Kim JY, Kim MH, Hong SP, Park BC. Characteristics of nonbalding scalp zones of androgenetic alopecia in East Asians. Clin Exp Dermatol. 2015 Apr 1;40(3):279-85.
- Kitamura A, Yoshimura M, Nakagami G, Yabunaka K, Sanada H. Changes of tissue images visualised by ultrasonography in the process of pressure ulcer occurrence. J Wound Care. 2019 Apr 1;28:S18-22.
- Miller IJ, Stapelberg M, Rosic N, Hudson J, Coxon P, Furness J, et al. Implementation of artificial intelligence for the detection of cutaneous melanoma within a primary care setting: prevalence and types of skin cancer in outdoor enthusiasts. PeerJ. 2023;11.
- Pham TC, Luong CM, Hoang VD, Doucet A. AI outperformed every dermatologist in dermoscopic melanoma diagnosis, using an optimized deep-CNN architecture with custom mini-batch logic and loss function. Sci Rep. 2021 Dec 1;11(1).
- Goyal M, Knackstedt T, Yan S, Hassanpour S. Artificial intelligence-based image classification methods for diagnosis of skin cancer: Challenges and opportunities. Comput Biol Med. 2020 Dec 1;127.
- Li CX, Fei WM, Shen CB, Wang ZY, Jing Y, Meng RS, et al. Diagnostic capacity of skin tumor artificial intelligence-assisted decision-making software in real-world clinical settings. Chin Med J (Engl). 2020 Sep 5;133(17):2020-6.
- Marka A, Carter JB, Toto E, Hassanpour S. Automated detection of nonmelanoma skin cancer using digital images: A systematic review. BMC Med Imaging. 2019 Feb 28;19(1).
- Zouridakis G, Wadhawan T, Situ N, Hu R, Yuan X, Lancaster K, et al. Melanoma and other skin lesion detection using smart handheld devices. Methods in Molecular Biology. 2015;1256:459-96.
Qualification of the Responsible Evaluators
Taig Mac Carthy
- Job position:
JD-003 Design and Development Manager
- Education: with a specialization in Strategic Management and Innovation from Copenhagen Business School, he has a foundational understanding of business practices essential in product development. His knowledge in quality management systems is well-established, having completed ISO 13485, ISO 9001:2015, and ISO 27001 Lead Auditor certifications from Bureau Veritas Group. These certifications underscore his ability to maintain high-quality standards in device manufacturing. Additionally, his training in ICH Good Clinical Practice and as an Equal Opportunity Agent, alongside courses in Python, Data Science, and Graphic Design, provide a diverse skill set applicable to his current role. His academic journey also includes a degree from the University of the Basque Country.
- Experience with the product/process/technology: solid background in both the medical and entrepreneurial fields. He has contributed to four scientific publications in computer vision applied to medicine, showcasing his expertise in areas directly relevant to medical device development. His involvement from the inception of the company, given his position as co-founder, has afforded him comprehensive knowledge of the device's development journey. His six years as a front-end software developer and the founding of three companies demonstrate his technical skills and entrepreneurial mindset. Additionally, his authorship of two business management books indicates his grasp on business operations, all of which collectively support his capacity to lead in design and development.
- Valuation: QUALIFIED
- Date: 02/07/2023
- Approved by: Alfonso Medela (
JD-005 Technical Manager & PRRC
)
Alfonso Medela
- Job position:
JD-005 Technical Manager & Person Responsible for Regulatory Compliance (PRRC)
- Education: He holds a degree in physics from the University of the Basque Country. In addition, he completed his training with a MSc. in Physics at the University of Groningen and a MSc. in Big Data and intelligence. University of Groningen and another Master in Big Data and Business Intelligence at the University of Deusto.
- Experience with the product/process/technology: expert in computer vision, machine learning and artificial intelligence with more than 5 years of experience in the development of projects with medical approaches. His experience includes his time at Tecnalia Research & Innovation where he worked as a data scientist focused on Deep Learning algorithms in the area of Computer Vision. He has written 7 papers on machine learning and image recognition, he also teaches workshops and courses on machine learning and deep learning. At the European level, he is one of the few experts on the few-shot learning methodology in the field of artificial intelligence.
- Training in risk management and other applicable: ISO 13485 and Medical Devices regulatory.
- Valuation: QUALIFIED
- Date: 02/07/2023
- Approved by: Andy Aguilar (
JD-001 General Manager
)
María Diez
- Job position:
JD-004 Quality manager & Person Responsible for Regulatory Compliance (PRRC)
- Education: María studied Biology at the Complutense University of Madrid. In addition, she holds a PhD on Biochemistry and Molecular Biology by the same University.
- Experience with the product/process/technology: With more than 7 years on Quality and Regulatory experience, María started developing her abilities implementing a Quality Management system based on ISO 15189, CLIA and Spanish sanitary regulations (specific for medical laboratories), clinical studies and in vitro Software as medical device. On her last work experience she developed and integrated QMS combining the ISO 9001, ISO 13485, ISO 15189 and ISO 27001 regulations with the requirements established at the 2017/746 European in vitro medical device regulations, again for a Software as medical device.
- Training in risk management and other applicable: ISO 14971, ISO 13485, ISO 9001, ISO 27001, 2017/745 Medical Device Regulations.
- Valuation: QUALIFIED
- Date: 02/07/2023
- Approved by: Alfonso Medela (
JD-005 Technical Manager & PRRC
)
Alberto Sabater
- Job position:
JD-009 Medical data scientist
- Education: Alberto studied Computer Engineering (data science specialization) at the University of Zaragoza, where he also obtained his PhD in Deep Learning and Computer Vision.
- Experience with the product/process/technology: PhD studies on Efficient scene understanding from video data have provided me extensive experience in:
- Many Computer Vision tasks (e.g. object detection, action recognition, semantic segmentation).
- The processing of different data modalities (e.g. RGB, text, point clouds, event and hyperspectral data).
- Neural Network design and implementation.
- Modern learning strategies (e.g. self-supervised, contrastive, weakly-supervised, multi-modal learning).
- Counts with publications in top-rated conferences and journals, as well as industry experience
- Training in risk management and other applicable: Not required.
- Valuation: QUALIFIED
- Date: 09/02/2023
- Approved by: Alfonso Medela (
JD-005 Technical Manager & PRRC
)
María Belén Hirigoity
- Job position:
- Dermatologist at Dermatological Institute Dr. Alonso
- Medical Advisor at Legit.Health
- Assistant University Professor at the School of Medicine UBA
- Education: Medical doctor degree from Universidad del Salvador, Buenos Aires, Argentina, with specializations in Dermatology from Bernardino Rivadavia Acute Care General Hospital. Master's degree in Advanced aesthetic and Laser techniques from the University of Buenos Aires (UBA) and Universidad Cardenal Herrera (CEU UCH), Valencia, Spain.
- Experience with the product/process/technology: With more than 8 years on the Dermatology field, she also has been given access to the device to test it by herself.
- Training in risk management and other applicable: Not required
- Valuation: QUALIFIED
- Date: 02/07/2023
- Approved by: Alfonso Medela (
JD-005 Technical Manager & PRRC
)
Constanza Balboni
- Job position:
- Specialist in Dermatology, Medical Aesthetics, Phlebology & Lymphology.
- Medical Advisor at Legit.Health
- Assistant University Professor on Dermatology at the University of Medicine UBA
- Education: Medical doctor degree at the Austral University Hospital. Specialization in Dermatology from Bernardino Rivadavia Hospital. Several specialization and degree in clinical dermatology, trichology, cosmetic dermatology and phlebology. Master's degree in Medical Aesthetics.
- Experience with the product/process/technology: With more than 6 years on the Dermatology field, she also has been given access to the device to test it by herself.
- Training in risk management and other applicable: Not required
- Valuation: QUALIFIED
- Date: 02/07/2023
- Approved by: Alfonso Medela (
JD-005 Technical Manager & PRRC
)
The clinical team CVs are documented as independent files:
- The internal employees CVs are archived within our Human Resources tool. The CVs are updated and revalidated yearly and archived as: YYYY_MM_Name_Surname_CV
- The external evaluators CVs are archived and saved in the corresponding evidences folder for the CER placed at GoogleDrive
Supporting documentation for the QMS
folder.
Additionally, the T-015-007 Declaration of interest Clinical evaluation team
of each evaluator is signed and placed at the TF Clinical evaluation folder
.
Clinical evaluation team justification
Selecting the right team for a Clinical Evaluation is a critical decision as the team's composition can greatly influence the quality and impartiality of the evaluation. The team's composition was carefully considered and selected to ensure a thorough and unbiased evaluation of the medical device.
JD-003
and JD-005
- Technical Expertise: The Technical Manager and Design and Development Manager, as founders of the company, have in-depth knowledge of the device's technical aspects. Their involvement ensures that the evaluation benefits from their comprehensive understanding of the device's design, development, and intended use. They can provide valuable insights into the device's technical performance and potential improvements.
- Ownership and Accountability: As founders, they have a vested interest in the device's success and safety. Their involvement underscores their commitment to the quality and safety of the product. Their accountability ensures that the evaluation process is conducted rigorously.
JD-004
- Regulatory Compliance and Quality Assurance: The Quality Manager and PRRC bring regulatory expertise to the evaluation process. Their presence is crucial for ensuring that the evaluation aligns with regulatory requirements and quality standards. They can verify that the device complies with essential regulatory and quality criteria.
- Risk Assessment and Mitigation: Their understanding of quality and regulatory compliance can help identify and address potential risks, contributing to the safety and performance of the device.
JD-009
- Data Analysis and Interpretation: The Medical Data Science Professional plays a key role in analyzing clinical data and drawing meaningful insights from it. Their expertise ensures that the data is evaluated objectively and rigorously, contributing to the credibility of the report.
- Evidence-Based Conclusions: Their involvement adds a layer of scientific rigor to the evaluation process, making sure that conclusions are evidence-based and supported by sound data analysis.
External Healthcare Professionals
- Independence and Objectivity: Including external healthcare professionals in the evaluation team enhances the objectivity and independence of the assessment. They provide an external perspective and reduce the potential for bias or conflicts of interest.
- Clinical Relevance: Healthcare professionals bring clinical expertise to the table, ensuring that the clinical benefits and safety of the device are assessed from a medical perspective.
Dates and signatures
Approval an acceptance of the present report by representing the manufacturer of the medical device covered by it:
Name | Position | Signature |
---|---|---|
Alfonso Medela | Technical Responsible | |
María Belén Hirigoity | Dermatologist. Medical Advisor at Legit.Health | |
Constanza Balboni | Dermatologist. Medical Advisor at Legit.Health |
Signature meaning
The signatures for the approval process of this document can be found in the verified commits at the repository for the QMS. As a reference, the team members who are expected to participate in this document and their roles in the approval process, as defined in Annex I Responsibility Matrix
of the GP-001
, are:
- Author: Team members involved
- Reviewer: JD-003, JD-004
- Approver: JD-005