ProMUSE: Progressive Multi-modal Uncertainty-guided Staged Evidential Alzheimer Disease Classification

Short summary

ProMUSE is a machine learning framework that intelligently decides when to use expensive medical imaging for Alzheimer's diagnosis. It starts with low-cost clinical assessments, estimates uncertainty, and only requests MRI or PET scans when necessary. On major clinical datasets, it maintains diagnostic accuracy while reducing imaging usage by 50-90%, potentially lowering barriers to early diagnosis.

• •Staged acquisition strategy: starts with clinical data, adds imaging only when uncertainty exceeds learned threshold
• •Reduces MRI/PET imaging costs by 50-90% while maintaining competitive accuracy on Alzheimer's datasets
• •Uses Dirichlet-based subjective logic and Dempster-Shafer theory for calibrated multimodal uncertainty quantification

Generated with AI, which can make mistakes.