Learning robust and task-invariant functional representation from fMRI through Siamese self-supervised learning

Jiyao Wang

doi:10.1016/j.media.2026.104221

Learning robust and task-invariant functional representation from fMRI through Siamese self-supervised learning

2026-07-23

0

OA

AI

DOI：10.1016/j.media.2026.104221

Original

Original request for help

Save

Abstract

En 中文

• A novel self-supervised representation learning framework for fMRI. • ROI-aligned augmentations and masking improve robustness. • Joint GNN–CNN encoder unifies ROI graphs and voxel-wise images. • Generalizes across diverse downstream prediction tasks. • Unified ROI-level interpretability on graph & image inputs.

Keywords:

ASD

fMRI

GNN

Contrastive learning

Learning robust and task-invariant functional representation from fMRI through Siamese self-supervised learning

Abstract

Journal

Organization