This research introduces a novel Concept Bottleneck Model approach for skin lesion diagnosis that improves transparency while maintaining diagnostic performance.

• Implements a two-step architecture that bases diagnoses on human-understandable medical concepts
• Reduces annotation burden through a unique learning process that infers concepts
• Achieves comparable accuracy to traditional black-box models while providing clear reasoning
• Enables medical professionals to understand and trust AI diagnostic decisions

This work addresses a critical barrier to AI adoption in healthcare by making diagnostic systems more transparent and accountable, potentially accelerating clinical integration of AI tools.

A Two-Step Concept-Based Approach for Enhanced Interpretability and Trust in Skin Lesion Diagnosis