Efficient Handwriting-Based Alzheimer,s Disease Diagnosis Using a Low-Rank Mixture of Experts Deep Learning Framework

Researchers propose LoRA-MoE, a parameter-efficient deep learning framework using low-rank adapters across specialized experts to diagnose Alzheimer's disease from handwriting patterns, reducing trainable parameters versus standard MoE while leveraging motor-cognitive degradation as a non-invasive digital biomarker.