Interpretable machine learning for dermatological disease detection: Bridging the gap between accuracy and...

Area of Science:

• Artificial Intelligence
• Medical Informatics
• Machine Learning

Background:

• Machine learning (ML) is crucial for prompt disease detection and treatment.
• ML models can be complex, necessitating understandable and trustworthy predictions.
• Skin disease detection is challenging due to symptom similarity, requiring high accuracy.

Purpose of the Study:

• To develop a highly accurate hybrid ML model for skin disease detection.
• To enhance the interpretability and trustworthiness of ML-based disease detection models.
• To apply Explainable AI (XAI) techniques for model insights.

Main Methods:

• A hybrid ML model combining Support Vector Machine (SVM) and XGBoost was developed.
• The model's performance was evaluated against existing ML models.
• Explainable AI frameworks, SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME), were explored.

Main Results:

• The proposed hybrid model achieved a superior accuracy of 99.26%.
• The model outperformed individual SVM, decision tree, and XGBoost models.
• SHAP and LIME provided local and global explanations for model predictions.

Conclusions:

• The hybrid SVM-XGBoost model offers a robust and accurate solution for skin disease detection.
• XAI techniques are valuable for building trust and understanding in diagnostic ML models.
• Accurate and interpretable ML models are essential for clinical applications in dermatology.