REQ_010 The device detects if the image is of clinical or dermatoscopic modality
Category​
Minor
Source​
- Alfonso Medela JD-005
INTERNAL SYSTEM INPUTS AND OUTPUTS ALARMS, WARNINGS AND MESSAGES USER DATABASE AND DATA DEFINITION ARCHITECTURE
Activities generated​
- MDS-452
Causes failure modes​
- The AI model is too sensitive, flagging images incorrectly as a different modality.
- The AI model is not sensitive enough, missing subtle differences between clinical and dermatoscopic images.
- The algorithm has been trained on a limited dataset, making it less effective at accurately distinguishing between clinical and dermatoscopic images.
- The algorithm might have biases that affect its ability to accurately identify the modality of diverse images.
- Notifications are not clear or specific enough for the user to understand the detected modality.
Related risks​
-
- Medical device input requirements are not defined to users to its proper operation
-
- Inadequate specification of the product intended purpose
-
- Inadequate specification of the product accessories
-
- Inaccurate training data: image datasets used in the development of the device are not properly labeled
-
- Biased or incomplete training data: image datasets used in the development of the device are not properly selected
-
- Stagnation of model performance: the AI/ML models of the device do not benefit from the potential improvement in performance that comes from re-training
-
- Degradation of model performance: automatic re-training of models decreases the performance of the device
User Requirement, Software Requirement Specification and Design requirement​
User Requirement 10.1: Users shall receive information regarding the modality (clinical/dermatoscopic) of the uploaded image.
Software Requirement Specification 10.2: Integrate algorithms capable of distinguishing between clinical and dermatoscopic image modalities.
Design Requirement 10.3: The device shall return modality information in a clear and standardized format, adhering to applicable healthcare data interchange standards.
Description​
The field of dermatology encompasses various image modalities, with the most common ones being clinical, dermatoscopic, ecographic, and histopathological images. It's worth noting that the latter two, ecographic and histopathological images, necessitate specialized equipment and expertise for interpretation, as they reveal internal tissue structures. On the other hand, clinical and dermatoscopic images are notably more accessible, enabling precise diagnoses in most cases. These images capture the external aspects of the skin or, in simpler terms, the dermis.
Understanding the modality of an image provides users with additional information, which can prove crucial for subsequent internal processing. This distinction implies the application of slightly different techniques when dealing with clinical or dermatoscopic images.
This requirement centers on the design and development of an algorithm for detecting image modalities, specifically clinical and dermatoscopic, which are the data types that underpin the functionality of our other algorithms.
Dataset preparation entails curating dermatoscopic images that cover a broad spectrum of pathologies, alongside clinical images. It's essential to emphasize that the dermatoscopic images will encompass a limited set of pathologies, primarily pigmented skin lesions. Therefore, it's imperative to represent these pathologies within the clinical modality as well. This prevents the misconception that pigmented lesions are exclusively found in dermatoscopic images.
Using this dataset, we will train a lightweight transformer model, placing a strong emphasis on execution speed. This is essential since this algorithm will operate continuously, independently of the specific functionality requested by the user.
Success metrics​
Goal | Metric | Target |
---|---|---|
Image modality is automatically detected | AUC | > 0.8 |
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:
- Tester: JD-017, JD-009, JD-005, JD-004
- Approver: JD-003