Development of an explainable machine learning model for Alzheimer's disease prediction using clinical and...

Area of Science:

• Computational neuroscience
• Medical informatics
• Machine learning in healthcare

Background:

• Early Alzheimer's disease (AD) detection is crucial for timely intervention.
• Existing diagnostic methods can be invasive or lack predictive power.
• Machine learning offers potential for data-driven early AD prediction.

Purpose of the Study:

• To present a reproducible machine learning methodology for early Alzheimer's disease prediction.
• To enhance model interpretability for clinical application.
• To develop a user-friendly tool for clinicians and researchers.

Main Methods:

• Utilized clinical and behavioral data for predictive modeling.
• Compared multiple classification algorithms, including Gradient Boosting.
• Integrated SHapley Additive exPlanations (SHAP) for feature interpretability.
• Developed an interactive web application using Streamlit.

Main Results:

• Gradient Boosting classifier achieved high performance (93.9% accuracy, 91.8% F1-score).
• SHAP analysis identified key predictive variables: MMSE, ADL, cholesterol, and functional scores.
• SHAP values provided global and individual feature contribution insights.
• The Streamlit application enabled real-time, explainable predictions.

Conclusions:

• The developed methodology provides a practical tool for early AD diagnosis and risk assessment.
• Explainable AI (SHAP) enhances clinical trust and understanding of predictive factors.
• The interactive web application facilitates informed clinical decision-making and personalized patient care.