Abstract
Cross-site generalization remains the primary barrier to deploying resting-state functional Magnetic Resonance Imaging (rs-fMRI)- based Autism Spectrum Disorder (ASD) classifiers in clinical settings. Graph transformer models trained on source acquisition sites achieve intra-site AUC above 0.82 on ABIDE-I but drop to 0.73 on fully unseen target sites, a gap driven by scanner-induced distribution shifts in functional connectivity matrices that existing domain adaptation methods cannot reliably suppress without erasing diagnostic signal. To address this limitation, we propose DANN-ALTER-Con, combining the Domain Adversarial Neural Network (DANN), the Long-range Brain Graph Transformer (ALTER) backbone, and Contrastive learning (Con), as a dual joint-learning framework for cross-site ASD classification. Specifically, we introduce the Gradient Reversal Layer Adversarial module (Grl-Adv), a site discriminator attached via a Gradient Reversal Layer that forces the ALTER encoder to produce site-invariant representations by reversing discriminator gradients during backpropagation, and the Supervised Contrastive Alignment branch (SupCon-Align), a supervised contrastive projection branch that actively clusters same-diagnostic-label embeddings across acquisition sites in the latent space. Evaluated on ABIDE-I under a strict site-disjoint protocol with 7 source sites for training, TRINITY site for model selection, and 11 fully unseen sites for testing, DANN-ALTER-Con achieves AUC 0.75 ± 0.08, outperforming the ALTER source-only baseline by 1.80% improvement and reducing cross-site AUC standard deviation from 0.09 to 0.08.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MWM_107.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=9CgJulraml
BibTex
@InProceedings{BazFat_AdversarialContrastive_MICCAISAT2026,
author = { Bazay, Fatima Ez-Zahraa AND Jennane, Rachid AND Barry, Martin AND Widmer, Antoine AND El Maliani, Ahmed Drissi},
title = { { Adversarial-Contrastive Domain Generalization for Cross-Site Autism Spectrum Disorder Classification from Resting-State fMRI } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17262},
month = {pending},
page = {pending}
}