dFCExpert: Learning Dynamic Functional Connectivity Patterns with Modularity and State Experts
Keywords: fMRI, dynamic brain functional connectome learning, brain modularity, state patterns, mixture of experts
Abstract: Modeling brain dynamic functional connectivity (dFC) patterns from functional Magnetic Resonance Imaging (fMRI) data is of paramount importance in neuroscience and medicine. Recently, many graph neural networks (GNN) models in conjunction with transformers or recurrent neural networks (RNNs) have been proposed and shown great potential for modeling dFC patterns in terms of pattern recognition and prediction performance. Although fruitful, several issues still hinder further performance improvement of these methods, such as neglecting the intrinsic brain modularity mechanism, and the interpretable state information of dFC patterns. To tackle these limitations, we propose dFCExpert to learn effective representations of dFC patterns in fMRI data with modularity experts and state experts. Particularly, using the GNN and mixture of experts (MoE), the modularity experts characterize the brain modularity organization in the graph learning process by optimizing multiple experts, with each expert capturing brain nodes with similar functions (in the same neurocognitive module); and the state experts aggregate temporal dFC features into a set of distinctive connectivity states by a soft prototype clustering methods, where the states can support different brain activities or are affected differently by brain disorders, thus revealing insights for interpretability. Experiments on two large-scale fMRI datasets demonstrate the superiority of our method over known alternatives, and the learned dFC representations show improved explainability and hold promise to improve clinical diagnosis.
Primary Area: applications to neuroscience & cognitive science
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 4935
Loading