Resting-state functional magnetic resonance imaging (rs-fMRI) is a noninvasive technique pivotal for understanding human neural mechanisms of intricate cognitive processes. Most rs-fMRI studies compute a single static functional connectivity matrix across brain regions of interest, or dynamic functional connectivity matrices with a sliding window approach. These approaches are at risk of oversimplifying brain dynamics and lack proper consideration of the goal at hand. While deep learning has gained substantial popularity for modeling complex relational data, its application to uncovering the spatiotemporal dynamics of the brain is still limited. In this study we propose a novel interpretable deep learning framework that learns goal-specific functional connectivity matrix directly from time…

5. Discussion and conclusion In this research study, we have demonstrated the effectiveness of our model, DSAM , in capturing the spatiodynamics of the brain and learning connectivity matrices relevant to prediction tasks. Our model leverages TCN, temporal attention, self-attention, and ROI-aware GNNs to comprehensively capture the spatiotemporal dynamics of brain connectivity directly from raw and noisy resting-state fMRI (rs-fMRI) data. Significantly, this approach enhances interpretability and offers a deeper understanding of the mechanisms underlying brain function. One of the key strengths of our model lies in its direct utilization of raw resting-state fMRI time points, bypassing the need for handcrafted FC features commonly employed by existing models such as BNT, FBNetGen, BrainRGIN, and BrainGNN. This approach represents a departure from conventional methods and presents a more challenging task of seamlessly integrating both temporal and spatial blocks, thus contributing to the model’s “heaviness”. Moreover, our model offers a significant advantage in its ability to learn directed goal-specific connectivity matrices. The directed FC matrices are different from correlation…

The Appreciated Gateway must be evaluated by a human keeper.
Does this declaration represent a genuine open research gap?