Generalizable Representation Learning for fMRI-based Neurological Disorder Identification

This repository contains the official implementation of MeTSK, a generalizable representation learning framework designed specifically for neurological disorder identification using functional MRI (fMRI) data.

Overview

MeTSK combines Meta-learning and Self-supervised learning to tackle challenges posed by scarce and heterogeneous clinical fMRI datasets. Our method achieves robust representation learning and significantly improves model generalization to unseen clinical data.

Paper published in Transactions on Machine Learning Research (TMLR):

• Generalizable Representation Learning for fMRI-based Neurological Disorder Identification

BibTeX citation:

@article{
cui2025generalizable,
title={Generalizable Representation Learning for f{MRI}-based Neurological Disorder Identification},
author={Wenhui Cui and Haleh Akrami and Anand Joshi and Richard Leahy},
journal={Transactions on Machine Learning Research},
issn={2835-8856},
year={2025},
url={https://openreview.net/forum?id=zF9IrMTjCC},
note={}
}