vSV-ViT: Variable-size SuperVertex Vision Transformer for Cortical Surface Learning in Alzheimer's Disease
The research addresses limitations of costly Alzheimer's diagnosis methods by proposing CSV-ViT, a Vision Transformer for structural MRI analysis. It introduces a novel method to partition the cortical surface into variable-sized, non-overlapping Supervertices (CSVs) that preserve regional information. This approach outperforms existing surface models in classifying AD diagnosis and pathology status using cortical thickness and curvature.