A new way to model individual brain communities for disease prediction

Researchers developed IC-GNN to map how brain community structures vary between people and across different diseases.

Researchers created IC-GNN to model individualized brain community organization. This model helps predict diseases by looking at how brain networks vary across individuals.

The model uses resting-state functional magnetic resonance imaging (fMRI) data. It forms distinct, individualized functional communities over atlas-defined brain regions. These communities are discrete and multiscale.

Brain networks show modular organization. Community structures change between individuals and during diseases. Individualized functional parcellations capture subject-specific variability in community boundaries.

IC-GNN refines community-guided connections. It adapts shared regional weights to each subject's community context. This allows for both individualized modeling and group-level comparisons.

The model was tested on more than 30,000 participants from ABIDE (autism spectrum disorder), ADNI (Alzheimer's disease), and the UK Biobank (depression). It achieved the highest performance compared with 15 baseline methods.

Individualized communities are reproducible within scans and across sessions. These communities carry predictive power on their own. They identify disease-related shifts in community organization.

Key facts

  • IC-GNN was tested on more than 30,000 participants from ABIDE, ADNI, and the UK Biobank.
  • The model achieves the highest performance compared with 15 baseline methods.
  • Individualized communities are reproducible within scans and across sessions.

All sections

Edition