A Privacy-Preserving Domain Adversarial Federated learning for multi-site brain functional connectivity analysis

Resting-state functional magnetic resonance imaging (rs-fMRI) and its derived functional connectivity networks (FCNs) have become critical for understanding neurological disorders. However, collaborative analyses and the generalizability of models still face significant challenges due to privacy regulations and the non-IID (non-independent and identically distributed) property of multiple data sources. To mitigate these difficulties, we propose Domain Adversarial Federated Learning (DAFed), a novel federated deep learning framework specifically designed for non-IID fMRI data analysis in multi-site settings. DAFed addresses these challenges through feature disentanglement, decomposing the latent feature space into domain-invariant and domain-specific components, to ensure robust global learning while preserving local data specificity. Furthermore, adversarial training facilitates effective knowledge transfer between labeled and unlabeled datasets, while a contrastive learning module enhances the global representation of domain-invariant features. We evaluated DAFed on the diagnosis of autism spectrum disorder (ASD) and further validated its generalizability in the classification of Alzheimer's disease (AD), demonstrating its superior classification accuracy compared to state-of-the-art methods. Additionally, an enhanced Score-CAM module identifies key brain regions and functional connectivity significantly associated with ASD and mild cognitive impairment (MCI), respectively, uncovering shared neurobiological patterns across sites. These findings highlight the potential of DAFed to advance multi-site collaborative research in neuroimaging while protecting data confidentiality. Introduction Resting-state functional magnetic resonance imaging (rs-fMRI) has emerged as a powerful and non-invasive technique for detecting abnormal brain activity [1]. Functional connectivity networks (FCNs), derived from rs-fMRI data, quantify temporal correlations between functional interactions in different brain regions, which are extensively utilized in studies of neurological disorders and mental illnesses [2, 3]. Recently, deep learning approaches have shown remarkable potential in analyzing fMRI data and FCNs, enabling significant breakthroughs in understanding brain function [4, 5]. Despite significant advancements in deep learning models, concerns over patient privacy and legal restrictions limit data sharing across institutions. This limitation poses challenges to the reproducibility and generalizability of data-driven approaches across diverse datasets [6, 7].