Abstract
Graph-based deep learning models, particularly graph neural networks and Transformer-based architectures, have gained increasing attention in functional brain network analysis. However, whether these models provide genuine advantages over simpler graph-free approaches remains unclear, partly due to inconsistent evaluation protocols across studies. We introduce \textbf{FBN-Bench}, a standardized \textbf{F}unctional \textbf{B}rain \textbf{N}etwork \textbf{Bench}mark for evaluating graph-based and graph-free models on openly available precomputed functional brain network datasets, including ABIDE and HCP-Gender. FBN-Bench provides fixed data splits, random seeds, and cross-validation protocols, enabling fair and reproducible comparison across graph-based and graph-free models. Through extensive experiments, we show that recently proposed graph-based models consistently fail to outperform strong graph-free baselines, such as logistic regression (LR), multilayer perceptron (MLP), and DeepSets, under identical evaluation conditions. Further representation analysis reveals that embeddings learned by graph-based deep learning models largely concentrate in low-dimensional subspaces, suggesting that the graph-structural inductive biases of current architectures may be poorly matched to these functional brain network datasets. We release \textbf{FBN-Bench} as an open-source repository to support reproducible evaluation, transparent model comparison, and future method development for functional brain network analysis.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/GRAIL_008.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=dQpPV1GQqz
BibTex
@InProceedings{GamGur_FBNBench_MICCAISAT2026,
author = { Gamgam, Gurur AND Yüksel Dal, Demet},
title = { { FBN-Bench: Do Graph-Based Deep Learning Models Improve Functional Brain Network Classification? } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17278},
month = {pending},
page = {pending}
}