CSV-ViT: A Vision Transformer with the Variable-sized Cortical Supervertices for Detection of Alzheimer's Disease Pathologies

2026-05-25

AI Summary

Structural MRI for Alzheimer’s disease (AD) pathology detection is limited by challenges in modeling cortical spherical topology, insufficient preservation of regions of interest (ROIs), and vertex redundancy caused by conventional uniform mesh partitioning. To address these issues, this work proposes an ROI-aware, vertex-level variable-size cortical super-vertex (CSV) partitioning scheme and introduces CSV-ViT, a visual Transformer architecture that accommodates variable patch sizes. CSV-ViT integrates mask-aware embedding and padding mechanisms to effectively process T1-weighted MRI data on the spherical cortical surface. Evaluated on three tasks—AD diagnosis, amyloid positivity, and tau positivity—CSV-ViT consistently outperforms existing surface-based models, demonstrating strong potential for MRI-based pre-screening of AD-related pathologies.

Abstract

Confirming Alzheimer's disease (AD) typically relies on positron emission tomography (PET), which remains costly and invasive, motivating the use of structural MRI-based prescreening. Deep learning on non-Euclidean manifolds, particularly brain cortical surfaces, faces significant challenges due to the data's spherical topology. Recent surface models have enabled learning from cortical surface data; however, imposing face-based uniform patches often causes duplicate vertices at patch boundaries. In general, many surface-based models are limited in their awareness of the region of interest (ROI), which can result in non-cortical regions, such as the medial wall, being included. We propose a cortical surface tokenization that performs ROI-preserving, vertex-based, variable-sized patch partitioning. We refer to these cortical surface patches as cortical supervertices (CSVs). Building on this representation, we design the CSV Vision Transformer (CSV-ViT), a variable-size patch-tolerant Vision Transformer that uses padding and a mask-aware patch embedding. We used T1-weighted MRI and evaluated our framework by classifying AD-related status into three categories: AD diagnosis, amyloid positivity, and tau positivity. Across the experiments, CSV-ViT achieved higher classification performance than recent surface-based models. The results suggest that the proposed CSV-ViT may support MRI-based prediction of AD-related status prior to PET or CSF confirmation.

Problem

• Alzheimer's disease
• cortical surface
• non-Euclidean manifold
• region of interest
• structural MRI

Innovation

• Cortical SuperVertices
• Variable-sized Patch Partitioning
• Mask-aware Patch Embedding
• Surface-based Vision Transformer
• ROI-preserving Tokenization