Multimodal Prediction of Alzheimer's Disease

This project, developed for Washington University's CSE 419A (AI for Health) course, explores early detection of Alzheimer's disease by combining two data modalities from the OASIS-1 dataset: brain MRI scans and clinical/demographic features. A convolutional neural network handles the imaging data, while classical machine learning models (scikit-learn, XGBoost) handle the clinical/tabular data, with predictions from both modalities combined into a final classifier.

The work is documented in a NeurIPS-format report with architecture diagrams for both the CNN and the combined classifier, and was presented as a final project demo for the course.

Highlights

• Multimodal fusion of MRI imaging and clinical/demographic data
• CNN for imaging analysis alongside scikit-learn/XGBoost models for tabular data
• Combined classifier integrating both modalities
• NeurIPS-format written report and presentation