Discovering Dynamic Functional Brain Networks via Spatial and Channel-wise Attention

TL;DR

This work introduces SCAAE, a spatial and channel-wise attention autoencoder that automatically discovers dynamic functional brain networks from fMRI without sliding windows. By embedding spatial-wise and channel-wise attention within a 3D CNN framework, it directly reconstructs and weights time-varying FBNs, avoiding linearity and independence constraints. Evaluated on the ADHD-200 rsfMRI dataset, SCAAE yields interpretable, temporally evolving FBNs and provides a novel view of brain-state transitions via IoU-based template mapping. The approach offers faster inference and a dynamic perspective for understanding brain networks, with potential extensions to task-based fMRI and biomarker discovery.

Abstract

Using deep learning models to recognize functional brain networks (FBNs) in functional magnetic resonance imaging (fMRI) has been attracting increasing interest recently. However, most existing work focuses on detecting static FBNs from entire fMRI signals, such as correlation-based functional connectivity. Sliding-window is a widely used strategy to capture the dynamics of FBNs, but it is still limited in representing intrinsic functional interactive dynamics at each time step. And the number of FBNs usually need to be set manually. More over, due to the complexity of dynamic interactions in brain, traditional linear and shallow models are insufficient in identifying complex and spatially overlapped FBNs across each time step. In this paper, we propose a novel Spatial and Channel-wise Attention Autoencoder (SCAAE) for discovering FBNs dynamically. The core idea of SCAAE is to apply attention mechanism to FBNs construction. Specifically, we designed two attention modules: 1) spatial-wise attention (SA) module to discover FBNs in the spatial domain and 2) a channel-wise attention (CA) module to weigh the channels for selecting the FBNs automatically. We evaluated our approach on ADHD200 dataset and our results indicate that the proposed SCAAE method can effectively recover the dynamic changes of the FBNs at each fMRI time step, without using sliding windows. More importantly, our proposed hybrid attention modules (SA and CA) do not enforce assumptions of linearity and independence as previous methods, and thus provide a novel approach to better understanding dynamic functional brain networks.

Figures (6)

• Figure 1: The key contribution of SCAAE is to construct FBNs via attention mechanism. A Channel-wise Attention (CA) module is used to weight the importance of FBNs for selecting FBNs automatically and a Spatial-wise attention (SA) module is used to capture the neural activities regions in the whole brain.
• Figure 2: The FBN templates constructed by ICA 3. The activated voxels are concentrated in the regions nearby themselves. The strong spatial correlation enables the attention mechanism to capture the features of neural activities.
• Figure 3: Overview of the FBNs derived from SCAAE. These are the outputs we obtain from the attention modules, and we can observe that the attention mechanism finds the activation regions perfectly.
• Figure 4: The method can observe changes in FBNs at each time step. These two figures show the state transition process of FBNs and can observe a gradual change process. This is what previous methods cannot provide.
• Figure 5: Comparison with SDL and ICA. The IoU refers to the spatial similarity of the corresponding methods and templates. The templates are based on ICA 3.