Personalized Federated Dictionary Learning for Modeling Heterogeneity in Multi-site fMRI Data
This work addresses data privacy and heterogeneity issues in multi-site neuroimaging studies, offering a federated learning solution that is incremental by building on existing dictionary learning and federated methods.
The paper tackles the problem of modeling heterogeneity in multi-site fMRI data under privacy constraints by proposing Personalized Federated Dictionary Learning (PFedDL), which decomposes site-specific dictionaries into shared global and personalized local components, resulting in improved accuracy and robustness on the ABIDE dataset compared to existing methods.
Data privacy constraints pose significant challenges for large-scale neuroimaging analysis, especially in multi-site functional magnetic resonance imaging (fMRI) studies, where site-specific heterogeneity leads to non-independent and identically distributed (non-IID) data. These factors hinder the development of generalizable models. To address these challenges, we propose Personalized Federated Dictionary Learning (PFedDL), a novel federated learning framework that enables collaborative modeling across sites without sharing raw data. PFedDL performs independent dictionary learning at each site, decomposing each site-specific dictionary into a shared global component and a personalized local component. The global atoms are updated via federated aggregation to promote cross-site consistency, while the local atoms are refined independently to capture site-specific variability, thereby enhancing downstream analysis. Experiments on the ABIDE dataset demonstrate that PFedDL outperforms existing methods in accuracy and robustness across non-IID datasets.