Alzheimer's Prediction via eXplainable AI and SHAP-Driven Feature Attribution

Alzheimer's disease (AD), a degenerative neurological condition characterized by memory loss and cognitive impairment, is a significant global health isuue. Timely intervention and better patient care can be facilitated by early and precise AD prediction. The SHapley Additive exPlanations (SHAP) interpretability model is used in this paper to offer a Machine Learning framework for Alzheimer's disease prediction. To categorize patients according to clinical and neuroimaging characteristics, the study combines many machine learning techniques like Random Forest, XGBoost, and Logistic Regression. SHAP is used to analyze feature contributions and improve the transparency of the model. Using a simulated dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI) framework, the proposed SHAP-augmented model performed better than baseline classifiers with an accuracy of 92.4% and an F1-score of 0.90. Explainable clinical insights are made possible by the interpretability of SHAP, which aids in identifying the most significant biomarkers, such as hippocampus volume, Mini-Tau levels in the cerebrospinal fluid (CSF) and the Mental State Examination (MMSE) score.