AI Model Trails Expert Skin Lesion Readers
A foundation artificial intelligence model surpassed less experienced physicians but did not outperform expert dermatologists in multiclass skin lesion diagnosis.
By
Andrea Surnit
June 27, 2026
Clinical Scorecard: AI Model Trails Expert Skin Lesion Readers
At a Glance
Category Detail
Condition Skin Lesions Key Mechanisms Comparison of AI systems and physician readers in diagnosing skin lesions using dermoscopy. Target Population Physicians with varying levels of dermoscopy experience. Care Setting Diagnostic study using retrospectively collected images.
Key Highlights
AI outperformed physicians with less than 3 years of experience but not those with over 10 years. Unimodal AI model achieved 72% accuracy, while expert physicians reached 74%. Multimodal AI model performed worse than unimodal despite additional clinical data. AI systems showed higher specificity but not higher multiclass diagnostic accuracy compared to expert readers. Study suggests AI may serve as a decision-support tool for less experienced clinicians. Guideline-Based Recommendations
Diagnosis
Use AI as a supplementary tool for diagnostic support in skin lesion evaluation. Management
Maintain active dermoscopy training for clinicians to prevent deskilling. Monitoring & Follow-up
Consider AI systems for systematic secondary review to reduce diagnostic errors. Risks
Overreliance on AI tools may lead to decreased diagnostic skills among clinicians. Patient & Prescribing Data
Patients with skin lesions requiring diagnosis.
AI tools may enhance diagnostic accuracy and confidence in less experienced clinicians.
Clinical Best Practices
Incorporate AI tools into training programs for dermatology trainees. Encourage collaboration between AI systems and expert clinicians for optimal outcomes. Related Resources & Content
