New federated learning framework improves brain connectivity analysis across sites

Researchers have developed FedDOSE, a novel federated learning framework designed to improve the analysis of dynamic functional connectivity in brain imaging data across multiple sites. This framework addresses the challenge of statistical heterogeneity inherent in multi-site datasets by explicitly decomposing site-specific effects. FedDOSE utilizes a Modularity-Guided Tucker Decomposition for high-dimensional dynamic functional connectivity tensors and employs Optimal Transport and Procrustes analysis for aligning class-specific prototypes. Experiments on datasets for Autism Spectrum Disorder and Attention-Deficit Hyperactivity Disorder demonstrate FedDOSE's superior performance in detection compared to existing methods. AI

IMPACT Enhances the ability to analyze complex, multi-site brain imaging data, potentially leading to more accurate diagnoses of neurological conditions.

RANK_REASON This is a research paper detailing a new framework for analyzing brain imaging data using federated learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• ABIDE-II
• ADHD-200
• attention deficit hyperactivity disorder
• autism
• FedDOSE
• federated learning
• functional magnetic resonance imaging
• Modularity-Guided Tucker Decomposition
• Optimal Transport
• Procrustes analysis

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →