Problem
Research questions and friction points this paper is trying to address.
structural-functional brain interaction
multimodal integration
connectome
cognitive function
graph representation
Learning Structural-Functional Brain Representations through Multi-Scale Adaptive Graph Attention for Cognitive Insight
🤖 AI Summary
This work proposes MAGNet, a novel framework for uncovering cognitive mechanisms by jointly modeling structural and functional brain connectivity. Integrating morphological features from structural MRI and functional connectivity from resting-state fMRI, MAGNet constructs a hybrid graph that captures both direct and indirect neural pathways. It introduces a multi-scale adaptive graph attention mechanism to enable end-to-end learning of structure–function interactions by leveraging local and global contextual information. A Transformer-based graph neural network, combined with a cross-modal joint loss, simultaneously optimizes modality consistency and task-specific prediction objectives. Experiments on the ABCD dataset demonstrate that MAGNet significantly outperforms existing baselines, validating its effectiveness in multimodal brain network modeling and cognitive function analysis.
📝 Abstract
Understanding how brain structure and function interact is key to explaining intelligence yet modeling them jointly is challenging as the structural and functional connectome capture complementary aspects of organization. We introduced Multi-scale Adaptive Graph Network (MAGNet), a Transformer-style graph neural network framework that adaptively learns structure-function interactions. MAGNet leverages source-based morphometry from structural MRI to extract inter-regional morphological features and fuses them with functional network connectivity from resting-state fMRI. A hybrid graph integrates direct and indirect pathways, while local-global attention refines connectivity importance and a joint loss simultaneously enforces cross-modal coherence and optimizes the prediction objective end-to-end. On the ABCD dataset, MAGNet outperformed relevant baselines, demonstrating effective multimodal integration for advancing our understanding of cognitive function.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Multi-scale Adaptive Graph Network
structure-function integration
graph attention
multimodal fusion
cognitive prediction
🔎 Similar Papers
No similar papers found.
