Keywords: Temporal Graph, Generative Model, Deep Learning, Functional Magnetic Resonance Imaging
Abstract: Spatiotemporal graphs are a natural representation of dynamic brain activity derived from functional magnetic imaging (fMRI) data. Previous works, however, tend to ignore time dynamics of the brain and focus on static graphs. In this paper, we propose a temporal graph deep generative model (TG-DGM) which clusters brain regions into communities that evolve over time. In particular, subject embeddings capture inter-subject variability and its impact on communities using neural networks. We validate our model on the UK Biobank data. Results of up to 0.81 AUC ROC on the task of biological sex classification demonstrate that injecting time dynamics in our model outperforms a static baseline.
Paper Type: both
Primary Subject Area: Unsupervised Learning and Representation Learning
Secondary Subject Area: Application: Other
Paper Status: original work, not submitted yet
Source Code Url: We are working on a full journal publication and would rather submit the code then.
Data Set Url: We are working on a full journal publication and would prefer to open source the data then.
Registration: I acknowledge that publication of this at MIDL and in the proceedings requires at least one of the authors to register and present the work during the conference.
Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.