Using Artificial Intelligence as a Diagnostic Decision Support Tool in Skin Disease: Protocol for an Observational...

Area of Science:

• Dermatology
• Artificial Intelligence
• Medical Informatics

Background:

• Dermatological conditions are prevalent, accounting for significant primary care visits and specialist referrals.
• Machine learning (ML) offers potential for analyzing complex data and supporting dermatological diagnostics, particularly in image analysis for skin cancer detection.

Purpose of the Study:

• To prospectively validate an image analysis ML model as a diagnostic decision support tool for dermatological conditions.

Main Methods:

• A prospective study involving 100 patients with skin problems presenting to general practitioners (GPs) in central Catalonia.
• Anonymized images of skin diseases were analyzed by an ML model (screening 44 conditions) and compared with teledermatology consultations and GP assessments.
• Performance metrics including precision, sensitivity, specificity, and accuracy were calculated for the ML model against dermatologist and GP diagnoses.

Main Results:

• Patient recruitment was completed within 5 months, with all images processed and analysis initiated.
• The study is currently analyzing results comparing ML model performance with expert diagnoses.

Conclusions:

• The study will elucidate the effectiveness and limitations of ML models in dermatological diagnostics.
• External validation is crucial for the regulatory approval and implementation of ML diagnostic systems in primary care settings.