multimodal-brain-connectivity-gnn

What problem does it solve?

Integrates fMRI, DTI, and sMRI data into a single, interpretable graph neural network to predict cognitive outcomes and reveal key multimodal biomarkers.

Core Features & Use Cases

• Multimodal integration: fMRI functional connectivity (FC), DTI structural connectivity (SC), and sMRI morphometric features are jointly modeled.
• Atlas-based parcellation: uses Glasser atlas to ensure cross-modality alignment across 360 cortical regions.
• Learnable connectivity weighting: edge masks learn to weight connections based on modality differences for interpretable results.
• Graph neural backbone: uses GAT/GCN layers to fuse region-level features and produce graph-level predictions.
• Explainability: provides modality-wise importance and important connections for biomarker discovery and clinical insights.
• Use cases: cognitive score prediction, biomarker discovery, disease profiling, and clinical prognosis in neuroimaging research.

Quick Start

Train the model on your fMRI, DTI, and sMRI data to predict cognitive outcomes and identify key multimodal biomarkers.