Jyothi Sanam

Alzheimer’s starts damaging the brain years before symptoms appear. Early detection could let patients start treatment when it actually helps, but current diagnostic methods catch the disease too late. This project builds a classifier that combines two types of brain MRI showing brain shrinkage and PET images revealing metabolic changes—to detect Alzheimer’s and Mild Cognitive Impairment. The system uses Transformers to fuse information from both scan types, letting the model learn which brain regions matter most for diagnosis. I’ll test whether this multimodal approach beats using MRI or PET alone. The classifier distinguishes four stages: cognitively normal, stable MCI, progressive MCI, and Alzheimer’s disease. This matters because progressive MCI patients convert to full Alzheimer’s within five years—they need early intervention most. The project delivers trained models, attention maps showing which brain regions the transformer focused on, and assessment of where the system fails.

TBD