Smart Health

Explainable AI for Alzheimer's

This project studies multimodal Alzheimer’s prediction with explainable AI so that predictions remain useful to practitioners rather than opaque black-box outputs.

Back to Research
Explainable AI for Alzheimer's project illustration.

Research Area

Smart Health

Key Methods

XAI, Multimodal ML, ADNI

Project Summary

Multimodal prediction with transparent decision support using SHAP, LIME, and case-based reasoning.

The research targets classification of patients into CN, LMCI, and AD categories using ADNI data while preserving transparency in how models arrive at predictions. Develop strong predictive models for Alzheimer’s progression while ensuring both local and global explainability. The modeling stack includes DT, RF, SVM, LR, XGBoost, and MLP, combined with Bayesian optimization and feature selection. Local XAI methods include LIME, SHAP, and case-based reasoning for patient-level insights. Global interpretability emphasizes feature importance and medically oriented embedding fusion for improved prediction.