Early Detection of Alzheimer's Disease Using Explainable Machine Learning on Clinical Biomarkers: A Multi-Class Classification Study Using the Alzheimer's Disease Neuroimaging Initiative (ADNI) Dataset

In three linesXGBoost model for three-class Alzheimer's detection (normal cognition, mild cognitive impairment, Alzheimer's) on 1,641 ADNI subjects. Macro AUC-ROC 0.983 in cross-validation, 0.982 on test set. SHAP analysis identifies CDR Global as dominant predictor for NC/MCI, CDR-SB and MMSE for AD.

Summary generated by Claude — human-verified