Multi-Scale Structural MRI Features Reveal Task-Based Functional Connectivity and Its Alterations in Psychiatric...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging

Background:

• Understanding the brain's structure-function relationship is key in neuroscience.
• Current models often use single-scale structural data, missing the brain's hierarchical organization.
• Integrating multi-scale structural features can improve predictions of functional activity.

Purpose of the Study:

• To develop a deep learning framework, CoGA-MTN, for predicting task-based functional connectivity (FC) using multi-scale structural MRI data.
• To jointly identify disease-related alterations in brain networks.
• To explore the relationship between multi-scale brain structure, function, and psychopathology.

Main Methods:

• Introduced the Collaborative Graph Attention Multi-Task Network (CoGA-MTN), a dual-branch graph attention network.
• Extracted global statistical and local topological features from structural MRI.
• Employed a cross-modal task-coordinated learning mechanism for FC prediction and multi-disease classification.

Main Results:

• CoGA-MTN outperformed single-scale models in task-conditioned FC prediction (PCC = 0.657 ± 0.02).
• Achieved macro F1 scores of 0.68-0.75 for classifying three psychiatric disorders.
• Reconstructed a stable whole-brain architecture conserved across tasks, revealing diagnosis-specific network discrepancies.

Conclusions:

• CoGA-MTN provides a unified framework for linking multi-scale brain structure, dynamic function, and psychopathology.
• The model accurately predicts functional connectivity and aids in diagnosing psychiatric disorders.
• This approach advances our understanding of brain organization and disease mechanisms.