Skip to main content
QMSQMS
QMS
  • Welcome to your QMS
  • Quality Manual
  • Procedures
  • Records
  • Legit.Health Plus Version 1.1.0.0
  • Legit.Health Plus Version 1.1.0.1
  • Legit.Health Utilities
  • Licenses and accreditations
  • Applicable Standards and Regulations
  • BSI Non-Conformities
    • Technical Review
      • Round 1
        • M1: Diagnostic Function
          • Q1: IFU Performance Claims
            • Question
            • Research and planning
            • Response
            • fyi
              • Why Performance Claims Are Not Condition-Specific: Clinical Evidence Rationale
              • The Device Provides an ICD-11 Distribution, Not a Diagnosis
              • Question for Jordi: Studies in the IFU bibliography that may raise concerns
          • Q2: Test Environment
          • Q3: Biofilm/Slough Verification
          • Q4: T377 Test Results
        • M2: Software V&V
        • N1: Information Supplied
        • N2: Usability
        • N3: Risk Management
    • BSI Non-Conformities
  • Pricing
  • Public tenders
  • BSI Non-Conformities
  • Technical Review
  • Round 1
  • M1: Diagnostic Function
  • Q1: IFU Performance Claims
  • fyi
  • The Device Provides an ICD-11 Distribution, Not a Diagnosis

The Device Provides an ICD-11 Distribution, Not a Diagnosis

Purpose of This Document​

This document catalogues every location in the Legit.Health Plus v1.1.0.0 technical documentation where the following regulatory argument is made -- either explicitly or implicitly:

The device does NOT diagnose specific conditions. Instead, it provides an interpretative distribution representation across ALL visible ICD-11 categories.

This argument is critical because it means the device's performance should not be evaluated against condition-specific sensitivity/specificity (as with diagnostic tests for disease X or Y), but rather against distributional accuracy metrics (top-k accuracy across all ICD-11 categories).

The argument appears 51 times across 20+ documents, touching every major area of the technical file.

The Core Regulatory Argument​

The argument rests on three pillars:

  1. The output is a probability distribution, not a binary result. The device outputs a normalised probability vector p = [p_1, p_2, ..., p_n] where sum(p_i) = 1, covering all visible ICD-11 categories. This is mathematically and conceptually different from a diagnostic test that returns positive/negative for a specific condition.

  2. The device covers ALL visible ICD-11 categories, not specific ones. The intended medical indication is "all diseases of the skin" -- not eczema, psoriasis, melanoma, or any particular condition. Performance evidence should therefore address the distribution as a whole.

  3. The device never confirms a diagnosis. It is a clinical decision support tool; the final diagnostic decision remains with the healthcare professional. The device output is one element among many in clinical assessment.

Catalogue of Occurrences​

1. Canonical Definition (Reusable Components)​

The single source of truth for the device's output definition lives in the reusable translations at packages/reusable/translations/en.json. Every document that renders the <Diagnosis />, <IntendedPurpose />, <WhatIs />, <WarningsDeviceOutput />, or <MedicalConditions /> components inherits this canonical phrasing.

1.1 The diagnosis.content / diagnosis.description string​

  • File: packages/reusable/translations/en.json (lines 118-136)
  • Type: EXPLICIT
  • Quote: "interpretative distribution representation of possible International Classification of Diseases (ICD) categories that might be represented in the pixels content of the image"
  • Significance: This is the single canonical definition. It propagates to every document that uses <Diagnosis />.

1.2 The whatIs.content string​

  • File: packages/reusable/translations/en.json (lines 130-132)
  • Type: IMPLICIT
  • Quote: "computational software-only medical device leveraging computer vision algorithms to process images"
  • Significance: Frames the device as a computational tool, not a diagnostic instrument.

2. Intended Purpose Statements​

2.1 Primary Intended Use statement​

  • File: packages/reusable/translations/en.json (lines 422-428)
  • Type: EXPLICIT
  • Quote (resolved template): "The device is a [computational software-only medical device], enhancing efficiency and accuracy of care delivery, by providing: an [interpretative distribution representation of possible ICD categories...]"
  • Rendered in: IntendedPurpose.mdx, STED, Device Description, GSPR, CEP, CER, and all clinical investigation protocols.

2.2 Recognition heading​

  • File: packages/reusable/translations/en.json (lines 442-448)
  • Type: EXPLICIT
  • Quote: Heading: "Image-based recognition of visible ICD categories" / Text: "The device is intended to provide an [interpretative distribution representation...]"

2.3 Device Usage Warning​

  • File: packages/reusable/translations/en.json (lines 462-464)
  • Type: EXPLICIT
  • Quote: "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."
  • Also appears verbatim in: STED (r-tf-sted.mdx, line 70)

2.4 Body Structures: "distribute the probabilities across ICD categories that are visible"​

  • File: packages/reusable/translations/en.json (lines 530-532)
  • Type: EXPLICIT
  • Quote: "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."

2.5 Intended Medical Indication: "all diseases of the skin"​

  • File: packages/reusable/translations/en.json (lines 562-564)
  • Type: EXPLICIT
  • Quote: "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 [...] the dermis, the cutaneous vasculature and the subcutaneous tissue."
  • Significance: Says "all diseases of the skin" -- not specific conditions.

3. Warnings -- Understanding the Nature of Device Output​

This is the most forceful and explicit articulation of the argument, rendered via the <WarningsDeviceOutput /> component.

3.1 The Distribution Explanation (single clearest statement)​

  • File: packages/reusable/translations/en.json (lines 570-572)
  • Type: EXPLICIT
  • Quote: "The device outputs a probabilistic distribution across all ICD-11 categories related to skin diseases. Unlike diagnostic tests (e.g., COVID-19 or HIV tests) that provide binary results (positive/negative), the device provides an array of ICD categories with distributed probabilities."

3.2 Cannot Confirm Condition​

  • File: packages/reusable/translations/en.json (lines 610-612)
  • Type: EXPLICIT
  • Quote: "The device cannot confirm the presence of a condition. This is a fundamental characteristic of its operation. The probabilistic distribution may be concentrated (leptokurtic) or dispersed (platykurtic) across categories, but the output is always a distribution across all ICD categories, never a binary result."

3.3 Comparison Table (Diagnostic Test vs. Device)​

  • File: packages/reusable/translations/en.json (lines 586-608)
  • Type: EXPLICIT
  • Content: A structured comparison table:
AspectDiagnostic TestLegit.Health Plus
Output typeBinary confirmationProbabilistic distribution
Output formatBoolean (TRUE/FALSE)Array of ICD categories and distributed probabilities

3.4 Clinical Decision Support framing​

  • File: packages/reusable/translations/en.json (lines 614-616)
  • Type: EXPLICIT
  • Quote: "The device serves as a clinical decision support tool that complements healthcare professionals' expertise. Diagnostic decisions remain with the healthcare professional who combines clinical knowledge, patient history, laboratory results, and other sources of information, including the device output."

4. Medical Conditions Component​

4.1 Distribution explanation in Medical Conditions​

  • File: packages/reusable/translations/en.json (lines 622-624)
  • Type: EXPLICIT
  • Quote: "the device does not provide a true positive or true negative for any specific condition, like diagnostic tests. On the contrary, the device provides the full list of ICD categories and the probability distributed across all the categories."
  • Significance: Directly negates condition-specific diagnostic framing.

5. Device Description and Specification Document (R-TF-Description-and-Specification)​

5.1 Output Specifications: "Probabilistic distribution across 239 ICD-11 categories"​

  • File: legit-health-plus-version-1-1-0-0/overview-and-device-description/r-tf-description-and-specification.mdx (line 281)
  • Type: EXPLICIT

5.2 ICD Category Recognition Module​

  • File: Same (line 210)
  • Type: EXPLICIT
  • Quote: "Vision Transformer (ViT) architecture for probabilistic ICD-11 classification"

5.3 Similar devices vs. comprehensive coverage​

  • File: Same (line 318)
  • Type: EXPLICIT
  • Quote: "most commercially available devices focus on specific conditions (particularly melanoma and skin cancer detection), whereas Legit.Health Plus offers comprehensive coverage of 239 ICD-11 categories"
  • Significance: Directly contrasts condition-specific competitors with this device's all-category approach.

5.4 Risk classification: "(not binary diagnostic confirmation)"​

  • File: Same (line 171)
  • Type: EXPLICIT
  • Quote: "The device provides [...] probabilistic distribution across ICD categories (not binary diagnostic confirmation for specific conditions)."

5.5 IMDRF 5.2.2: "purpose is not to determine whether or not that one time-critical condition is present"​

  • File: Same (line 179)
  • Type: EXPLICIT
  • Quote: "the device's purpose is not to determine whether or not that one time-critical condition is present (positive / negative), but to provide distributions of all ICD categories."
  • Significance: One of the strongest formulations in the entire file.

5.6 MDCG 2019-11: "does not treat nor diagnose"​

  • File: Same (line 137)
  • Type: EXPLICIT
  • Quote: "the device drives clinical management, but does not treat nor diagnose."

5.7 Brazilian regulation: "does not replace established gold standard techniques"​

  • File: Same (line 341)
  • Type: EXPLICIT
  • Quote: "The device provides diagnostic support but does not replace established gold standard techniques (e.g., biopsies for melanoma). Final diagnostic decisions remain with healthcare professionals."

6. STED Document (R-TF-STED)​

6.1 Critical Usage Warning​

  • File: r-tf-sted.mdx (lines 69-71)
  • Type: EXPLICIT
  • Quote: "The device should never be used to confirm a clinical diagnosis." (Verbatim repeat of Finding 2.3)

6.2 Novel Features: "clinical decision support tool that complements... without replacing clinical judgment"​

  • File: r-tf-sted.mdx (lines 259-261)
  • Type: EXPLICIT

7. AI/ML Description (R-TF-028-001)​

7.1 Clinical Models definition: "Interpretative distribution of ICD categories"​

  • File: r-tf-028-001-aiml-description.mdx (lines 36-37)
  • Type: EXPLICIT
  • Quote: Clinical models provide: "Interpretative distribution of ICD categories (diagnostic support for skin conditions)"

7.2 Clinical vs. Non-Clinical distinction​

  • File: Same (lines 53-60)
  • Type: EXPLICIT
  • Quote: Clinical models "directly fulfill the intended purpose: 'to provide quantitative data on clinical signs and an interpretative distribution of ICD categories to healthcare professionals'"

7.3 Mathematical definition of the output​

  • File: Same (lines 98-113)
  • Type: EXPLICIT
  • Quote: "this model outputs a normalized probability vector: p = [p_1, p_2, ..., p_n] where each p_i corresponds to the probability that the lesion belongs to the i-th ICD-11 category, and sum(p_i) = 1."
  • Significance: The mathematical definition makes the distributional nature unambiguous.

7.4 Performance metrics are distributional (Top-k), not condition-specific​

  • File: Same (lines 170-180)
  • Type: IMPLICIT
  • Observation: Performance thresholds use Top-1 >= 50%, Top-3 >= 60%, Top-5 >= 70% -- not sensitivity/specificity for individual conditions.
  • Significance: The choice of performance metrics directly reflects the distributional output model. A diagnostic device would use sensitivity/specificity per condition.

8. AI Development Plan (R-TF-028-002)​

8.1 Algorithm objective: "ranked probability distribution across relevant ICD-11 categories"​

  • File: r-tf-028-002-aiml-development-plan.mdx (lines 33-34)
  • Type: EXPLICIT

8.2 Objective: "assist clinicians in formulating a differential assessment"​

  • File: Same (line 41)
  • Type: IMPLICIT
  • Significance: Uses "differential assessment" (distributional) rather than "diagnosis" (definitive).

8.3 Clinical models: "interpretative distributions of ICD categories"​

  • File: Same (line 283)
  • Type: EXPLICIT

9. AI Development Report (R-TF-028-005)​

9.1 Overview: "Provides interpretative distribution of ICD-11 categories"​

  • File: r-tf-028-005-development-report.mdx (lines 21-23)
  • Type: EXPLICIT

9.2 Visible Category Consolidation​

  • File: Same (lines 104-120)
  • Type: IMPLICIT
  • Quote: "Grouping ICD-11 codes that cannot be reliably distinguished based on visual features alone into unified 'Visible ICD-11' categories."
  • Significance: Explains why the distribution is over "visible" categories -- reinforcing coverage of all visually distinguishable conditions rather than targeting specific ones.

9.3 Model definition: "probability distribution across ICD-11 disease categories"​

  • File: Same (lines 131-141)
  • Type: EXPLICIT

10. AI Release Report (R-TF-028-006)​

10.1 Clinical models: "interpretative distribution of ICD categories"​

  • File: r-tf-028-006-aiml-release-report.mdx (lines 15-16)
  • Type: EXPLICIT
  • Quote: "54 Clinical Models: Directly fulfill the device's intended purpose of providing quantitative data on clinical signs and interpretative distribution of ICD categories"

10.2 Output specification: "346 validated ICD-11 categories"​

  • File: Same (lines 89-113)
  • Type: EXPLICIT
  • Quote: "The ICD Category Distribution algorithm analyzes a dermatological image [...] to provide an interpretative distribution of probabilities across 346 validated ICD-11 categories."
  • Technical detail: Output type = "Probability distribution vector", Dimension = "346 values (one per validated ICD-11 category)", Constraint = "sum of all values = 1.0"

11. Clinical Evaluation Plan (R-TF-015-001)​

11.1 Mode of action: "interpretative distribution representation"​

  • File: R-TF-015-001-Clinical-Evaluation-Plan.mdx (line 340)
  • Type: EXPLICIT
  • Quote: "when the device is fed an image or a set of images, it outputs an interpretative distribution representation of possible ICD categories that might be represented in the pixels content of the image."

11.2 Equivalence assessment: "Not intended for use [...] as a standalone diagnostic tool"​

  • File: Same (line 506)
  • Type: EXPLICIT

12. Clinical Evaluation Report (R-TF-015-003)​

12.1 Mode of action: "interpretative distribution representation"​

  • File: R-TF-015-003-Clinical-Evaluation-Report.mdx (line 291)
  • Type: EXPLICIT
  • Quote: Identical to CEP Finding 11.1.

13. State of the Art (R-TF-015-011)​

13.1 Device description: "interpretative distribution representation"​

  • File: R-TF-015-011-State-of-the-Art.mdx (lines 11-14)
  • Type: EXPLICIT

13.2 Medical conditions: "all skin diseases listed and described in the ICD-11"​

  • File: Same (lines 73-75)
  • Type: EXPLICIT
  • Quote: "the medical conditions identified are all skin diseases listed and described in the ICD-11 (code 14)."
  • Significance: Explicitly says "all skin diseases" -- not specific conditions.

14. PMCF Report (R-TF-007-005)​

14.1 IFU claims interpretation​

  • File: R-TF-007-005-Post-Market-Clinical-Follow-up-PMCF-Report.mdx (lines 110, 127)
  • Type: EXPLICIT
  • Quote: Repeats the canonical Intended Purpose phrasing including "interpretative distribution representation of possible ICD classes."

15. Risk Management Report (R-TF-013-003)​

15.1 "Diagnostic support software that does not directly interact with patients"​

  • File: R-TF-013-003-Risk-Management-Report.mdx (line 110)
  • Type: IMPLICIT
  • Quote: "as a diagnostic support software that does not directly touch or interact with patients, P2 is always equal to 1. The device cannot directly cause physical harm; any harm pathway is indirect through clinical decision-making."
  • Significance: Frames as "diagnostic support" (not "diagnostic device") and establishes harm is only indirect.

16. Annex C Checklist (R-TF-013-005)​

16.1 "Is the medical device interpretive?" -- Yes, but "differential diagnoses"​

  • File: R-TF-013-005-Checklist-Annex-C.mdx (line 41)
  • Type: IMPLICIT
  • Quote: "It provides differential diagnoses, severity scores, and clinical recommendations. Final clinical decisions remain with the healthcare professional."
  • Significance: Uses "differential diagnoses" (distributional, plural) rather than "diagnosis" (definitive, singular).

17. Software Architecture Description (R-TF-012-029)​

17.1 Diagnosis support workflow: "probability distribution across ICD diagnostic categories"​

  • File: R-TF-012-029-Software-Architecture-Description.mdx (lines 58-59, 91-95)
  • Type: EXPLICIT

17.2 Expert table: "Generates a probability distribution across ICD categories"​

  • File: Same (line 125)
  • Type: EXPLICIT

18. Summative Evaluation Report (R-TF-025-007)​

18.1 Usability study: HCPs understanding of non-diagnostic nature​

  • File: R-TF-025-007-Summative-Evaluation-Report.mdx (lines 291-314)
  • Type: EXPLICIT
  • Key quotes from participants:
    • UP-02: "mas que un diagnostico concreto, una serie de diagnosticos diferenciales" (more than a concrete diagnosis, a series of differential diagnoses)
    • UP-05: Self-corrected from "diagnostico" (diagnosis) to "orientarnos" (guide us)
  • Result: 94.4% of participants (17/18) demonstrated correct understanding that the device is NOT a diagnostic tool.
  • Significance: Real-world validation that users understand the distributional nature of the output.

18.2 Residual risk: "probabilistic outputs"​

  • File: Same (line 308)
  • Type: EXPLICIT
  • Quote: "The device's architecture as a clinical decision support tool (requiring physician interpretation of probabilistic outputs) provides an inherent safety net."

19. Summative Evaluation Protocol (R-TF-025-004)​

19.1 Software Requirements: "Generate an aggregated ICD probability distribution"​

  • File: R-TF-025-004-Summative-Evaluation-Protocol.mdx (lines 76-77)
  • Type: EXPLICIT
  • Quote: "SRS-Q3Q: Generate an aggregated ICD probability distribution from a set of images"

20. AI/ML Risk Assessment (R-TF-028-011)​

20.1 Risk framing: "ICD category misclassification" (not "misdiagnosis")​

  • File: r-tf-028-011-aiml-risk-assessment.mdx (lines 28, 127-128)
  • Type: IMPLICIT
  • Significance: Even the risk vocabulary uses "misclassification" (distributional error) rather than "misdiagnosis" (diagnostic error).

20.2 Benefit-risk: "without replacing clinical judgment"​

  • File: Same (lines 141-143)
  • Type: IMPLICIT

21. AI Design Checks (R-TF-028-009)​

21.1 Purpose check: "ICD-11 probability distributions"​

  • File: r-tf-028-009-aiml-design-checks.mdx (lines 26-29)
  • Type: EXPLICIT
  • Quote: "The AI Description clearly defines the purpose: to provide clinical decision support through ICD-11 probability distributions"

22. Novel Features Component​

22.1 "Model calibration using temperature scaling for interpretable probability distributions"​

  • File: packages/reusable/translations/en.json (lines 830-832)
  • Type: IMPLICIT
  • Significance: Even design novelties are framed in distributional terms.

23. CAPA Plan (BSI CE Mark Closeout)​

23.1 BSI's own framing: "ICD class distribution"​

  • File: capa-plan-bsi-ce-mark-closeout.mdx (line 46)
  • Type: IMPLICIT
  • Quote: "Evidence that the device achieves the performance intended by the manufacturer for the diagnostic features described in [...] REQ_004 (ICD class distribution) cannot be confirmed"
  • Significance: Even BSI's non-conformity statement uses "ICD class distribution" (not "diagnosis"), reflecting the manufacturer's framing.

Summary​

Coverage by Document Area​

Document AreaFindingsExplicitness
Reusable translations (canonical definitions)12Almost all EXPLICIT
Device Description & Specification7All EXPLICIT
STED2Both EXPLICIT
AI/ML Description (028-001)4All EXPLICIT
AI Development Plan (028-002)3Mixed
AI Development Report (028-005)3All EXPLICIT
AI Release Report (028-006)2Both EXPLICIT
Clinical Evaluation Plan (015-001)2Both EXPLICIT
Clinical Evaluation Report (015-003)1EXPLICIT
State of the Art (015-011)2Both EXPLICIT
PMCF Report (007-005)1EXPLICIT
Risk Management Report (013-003)1IMPLICIT
Annex C Checklist (013-005)1IMPLICIT
AI Risk Assessment (028-011)2IMPLICIT
AI Design Checks (028-009)1EXPLICIT
Software Architecture (012-029)2Both EXPLICIT
Summative Evaluation Report (025-007)2Both EXPLICIT
Summative Evaluation Protocol (025-004)1EXPLICIT
Novel Features1IMPLICIT
CAPA Plan (BSI)1IMPLICIT
Total51

Three Strongest Formulations​

These are the locations where the argument is made most forcefully and could be cited directly:

  1. WarningsDeviceOutput component (Findings 3.1-3.4): Provides a comparison table, uses the COVID-19/HIV analogy, and states "the output is always a distribution across all ICD categories, never a binary result."

  2. MedicalConditions component (Finding 4.1): "the device does not provide a true positive or true negative for any specific condition [...] the device provides the full list of ICD categories and the probability distributed across all the categories."

  3. Device Description, IMDRF 5.2.2 (Finding 5.5): "the device's purpose is not to determine whether or not that one time-critical condition is present (positive / negative), but to provide distributions of all ICD categories."

Implications for Performance Evidence​

Because the device outputs a probability distribution across all ICD-11 categories rather than a binary diagnostic result for any specific condition:

  • Appropriate metrics: Top-k accuracy (top-1, top-3, top-5), calibration metrics, distributional accuracy, ranked probability scores
  • Inappropriate metrics: Condition-specific sensitivity/specificity, PPV/NPV for individual diseases, binary diagnostic accuracy for condition X or Y
  • Evidence strategy: Performance evidence should address the distribution as a whole (e.g., "the correct ICD category appears in the top-5 in 70% of cases"), not prove the device can diagnose specific conditions (e.g., "the device detects melanoma with 95% sensitivity")

This distinction is not merely semantic -- it is embedded in the mathematical definition of the output (a normalised probability vector summing to 1.0 across all categories), the intended use statements, the risk classification justification, the performance acceptance criteria, the usability validation, and even the notified body's own framing of the non-conformity.

Previous
Why Performance Claims Are Not Condition-Specific: Clinical Evidence Rationale
Next
Question for Jordi: Studies in the IFU bibliography that may raise concerns
  • Purpose of This Document
  • The Core Regulatory Argument
  • Catalogue of Occurrences
    • 1. Canonical Definition (Reusable Components)
      • 1.1 The diagnosis.content / diagnosis.description string
      • 1.2 The whatIs.content string
    • 2. Intended Purpose Statements
      • 2.1 Primary Intended Use statement
      • 2.2 Recognition heading
      • 2.3 Device Usage Warning
      • 2.4 Body Structures: "distribute the probabilities across ICD categories that are visible"
      • 2.5 Intended Medical Indication: "all diseases of the skin"
    • 3. Warnings -- Understanding the Nature of Device Output
      • 3.1 The Distribution Explanation (single clearest statement)
      • 3.2 Cannot Confirm Condition
      • 3.3 Comparison Table (Diagnostic Test vs. Device)
      • 3.4 Clinical Decision Support framing
    • 4. Medical Conditions Component
      • 4.1 Distribution explanation in Medical Conditions
    • 5. Device Description and Specification Document (R-TF-Description-and-Specification)
      • 5.1 Output Specifications: "Probabilistic distribution across 239 ICD-11 categories"
      • 5.2 ICD Category Recognition Module
      • 5.3 Similar devices vs. comprehensive coverage
      • 5.4 Risk classification: "(not binary diagnostic confirmation)"
      • 5.5 IMDRF 5.2.2: "purpose is not to determine whether or not that one time-critical condition is present"
      • 5.6 MDCG 2019-11: "does not treat nor diagnose"
      • 5.7 Brazilian regulation: "does not replace established gold standard techniques"
    • 6. STED Document (R-TF-STED)
      • 6.1 Critical Usage Warning
      • 6.2 Novel Features: "clinical decision support tool that complements... without replacing clinical judgment"
    • 7. AI/ML Description (R-TF-028-001)
      • 7.1 Clinical Models definition: "Interpretative distribution of ICD categories"
      • 7.2 Clinical vs. Non-Clinical distinction
      • 7.3 Mathematical definition of the output
      • 7.4 Performance metrics are distributional (Top-k), not condition-specific
    • 8. AI Development Plan (R-TF-028-002)
      • 8.1 Algorithm objective: "ranked probability distribution across relevant ICD-11 categories"
      • 8.2 Objective: "assist clinicians in formulating a differential assessment"
      • 8.3 Clinical models: "interpretative distributions of ICD categories"
    • 9. AI Development Report (R-TF-028-005)
      • 9.1 Overview: "Provides interpretative distribution of ICD-11 categories"
      • 9.2 Visible Category Consolidation
      • 9.3 Model definition: "probability distribution across ICD-11 disease categories"
    • 10. AI Release Report (R-TF-028-006)
      • 10.1 Clinical models: "interpretative distribution of ICD categories"
      • 10.2 Output specification: "346 validated ICD-11 categories"
    • 11. Clinical Evaluation Plan (R-TF-015-001)
      • 11.1 Mode of action: "interpretative distribution representation"
      • 11.2 Equivalence assessment: "Not intended for use [...] as a standalone diagnostic tool"
    • 12. Clinical Evaluation Report (R-TF-015-003)
      • 12.1 Mode of action: "interpretative distribution representation"
    • 13. State of the Art (R-TF-015-011)
      • 13.1 Device description: "interpretative distribution representation"
      • 13.2 Medical conditions: "all skin diseases listed and described in the ICD-11"
    • 14. PMCF Report (R-TF-007-005)
      • 14.1 IFU claims interpretation
    • 15. Risk Management Report (R-TF-013-003)
      • 15.1 "Diagnostic support software that does not directly interact with patients"
    • 16. Annex C Checklist (R-TF-013-005)
      • 16.1 "Is the medical device interpretive?" -- Yes, but "differential diagnoses"
    • 17. Software Architecture Description (R-TF-012-029)
      • 17.1 Diagnosis support workflow: "probability distribution across ICD diagnostic categories"
      • 17.2 Expert table: "Generates a probability distribution across ICD categories"
    • 18. Summative Evaluation Report (R-TF-025-007)
      • 18.1 Usability study: HCPs understanding of non-diagnostic nature
      • 18.2 Residual risk: "probabilistic outputs"
    • 19. Summative Evaluation Protocol (R-TF-025-004)
      • 19.1 Software Requirements: "Generate an aggregated ICD probability distribution"
    • 20. AI/ML Risk Assessment (R-TF-028-011)
      • 20.1 Risk framing: "ICD category misclassification" (not "misdiagnosis")
      • 20.2 Benefit-risk: "without replacing clinical judgment"
    • 21. AI Design Checks (R-TF-028-009)
      • 21.1 Purpose check: "ICD-11 probability distributions"
    • 22. Novel Features Component
      • 22.1 "Model calibration using temperature scaling for interpretable probability distributions"
    • 23. CAPA Plan (BSI CE Mark Closeout)
      • 23.1 BSI's own framing: "ICD class distribution"
  • Summary
    • Coverage by Document Area
    • Three Strongest Formulations
    • Implications for Performance Evidence
All the information contained in this QMS is confidential. The recipient agrees not to transmit or reproduce the information, neither by himself nor by third parties, through whichever means, without obtaining the prior written permission of Legit.Health (AI Labs Group S.L.)