Explainable machine learning models for early Alzheimer's disease detection using multimodal clinical data.

Area of Science:

• Artificial Intelligence in Medicine
• Biomedical Data Science
• Computational Neuroscience

Background:

• Alzheimer's disease (AD) poses a significant global health challenge, necessitating early and accurate prediction for effective intervention.
• Current machine learning models for AD prediction lack transparency, hindering clinical adoption due to their 'black-box' nature.

Purpose of the Study:

• To develop and evaluate explainable artificial intelligence (XAI) frameworks for AD prediction using multimodal patient data.
• To enhance model interpretability in AD prediction using SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) techniques.

Main Methods:

• Utilized a comprehensive dataset of 2,149 patients (aged 60-90) with demographic, medical, lifestyle, clinical, and cognitive data.
• Applied data preprocessing including MinMax normalization, SMOTE for class imbalance, and Backward Elimination Feature Selection.
• Evaluated six machine learning models, optimizing Random Forest (RF) with Ant Colony Optimization, and employing SHAP and LIME for interpretability.

Main Results:

• The optimized Random Forest model achieved high performance: 95% accuracy, 95% precision, 94% recall, 94% F1-score, and 98% AUC.
• SHAP analysis identified functional assessment, activities of daily living (ADL), memory complaints, and Mini-Mental State Examination (MMSE) as key predictors.
• LIME provided local explanations, confirming the clinical relevance of the identified features.

Conclusions:

• Integrating XAI with machine learning enhances transparency and trust in AI-driven AD diagnostic tools while maintaining high predictive accuracy.
• Future research should focus on external validation, clinical workflow integration, and addressing computational demands for real-world deployment.