SkinSage XAI: An explainable deep learning solution for skin lesion diagnosis

Area of Science:

• Dermatology
• Artificial Intelligence
• Medical Imaging

Background:

• Skin cancer is a major global health concern requiring early detection.
• Deep learning improves skin lesion classification but often lacks transparency.
• The
• black-box
• nature of AI models hinders clinical adoption.

Purpose of the Study:

• To develop an interpretable AI system for skin lesion categorization.
• To enhance diagnostic accuracy and transparency in dermatology.
• To support dermatologists in clinical decision-making.

Main Methods:

• Utilized explainable artificial intelligence (XAI) techniques for skin lesion classification.
• Employed a diverse dataset of approximately 50,000 skin lesion images (Customized HAM10000).
• Leveraged the Inception v3 model with Gradient-weighted Class Activation Mapping (Grad-CAM) and Local Interpretable Model-agnostic Explanations (LIME) for visual interpretability.

Main Results:

• SkinSage XAI accurately categorized seven types of skin lesions.
• Achieved high performance metrics: 96% accuracy, 96.42% precision, 96.28% recall, 96.14% F1-score, and 99.83% AUC.
• Provided clear visual explanations for model predictions.

Conclusions:

• SkinSage XAI offers a significant advancement in AI for dermatology, balancing accuracy and explainability.
• The system provides transparent and reliable diagnoses.
• Aims to improve dermatologists' decision-making and patient care.