GABIC

Generative Anatomically-Constrained Bidirectional Connectivity

MATLAB implementation of the GABIC framework for inferring directional effective connectivity using whole-brain Hopf models constrained by structural connectomes

This repository contains the MATLAB code used in:

Deco, G., Vidaurre, D., & Kringelbach, M. L. (2021). Revisiting the global workspace orchestrating the hierarchical organization of the human brain. Nature Human Behaviour, 5, 497–511. https://doi.org/10.1038/s41562-020-01003-6

This work introduces the GABIC framework, which estimates generative bidirectional effective connectivity by optimizing a whole-brain Hopf model to reproduce the empirical normalized directed transfer entropy (NDTE) flow. The method provides a model-based inference of causally directed interactions constrained by structural connectivity and optimized using particle swarm algorithms.

---

Purpose & Scope

• Estimate directed effective (generative) connectivity matrices from empirical NDTE using anatomically constrained models.
• Fit whole-brain Hopf oscillators with asymmetric coupling weights reflecting NDTE flow.
• Investigate the causal role of high-hierarchy brain regions (global workspace) through model-based lesioning.
• Provide a neurobiologically plausible simulation platform to reproduce empirical hierarchy of information flow.

---

Repository Structure

• Core MATLAB scripts implementing are in the folders:
  • MODEL: Whole-brain Hopf model dynamics.
  • NDTE: NDTE flow computations.

Note: Input neuroimaging data (fMRI and SC matrices) are not provided in the repository. Users should provide their own data.

---

Requirements

• MATLAB R2020a or later
• Required Toolboxes:
  • Optimization Toolbox (for particle swarm optimization)
  • Signal Processing Toolbox

---

Usage Instructions

1. Prepare input data:

   • Empirical NDTE flow matrix computed from preprocessed fMRI time series.
   • Structural connectivity (SC) matrix derived from DTI-based tractography.
   • Parcellation scheme (e.g., DK80) consistent across modalities.

2. Organize your data:

   • Store matrices as .mat files (e.g., NDTE_empirical.mat, SC.mat) within the data/ folder.
   • Ensure dimensional consistency across inputs (same number and order of regions).

3. Configure model parameters:

   • Set simulation parameters (e.g., number of particles, time step, duration) in the provided scripts or via custom configuration files.

4. Run GABIC optimization:

   • Launch the optimization routine to fit the generative model to the empirical NDTE matrix using particle swarm methods.

5. Analyze results:

   • Assess fit quality between empirical and simulated NDTE.
   • Examine the inferred asymmetric GABIC matrix.
   • Conduct in silico lesion analyses on selected regions to test causal influence on whole-brain dynamics.

6. Visualize outputs:

   • Use plotting functions to inspect connectivity matrices, node hierarchies, and simulation behaviour.

---

Data Sources & Ethics

• This repository does not include human subject data.

---

Citation

If you use this repository or adapt the GABIC methodology, please cite:

Deco, G., Vidaurre, D., & Kringelbach, M. L. (2021). Revisiting the global workspace orchestrating the hierarchical organization of the human brain. Nature Human Behaviour, 5, 497–511. https://doi.org/10.1038/s41562-020-01003-6

---

License

This project is licensed under the Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0).

---

Contact

For scientific questions, please contact:

Prof. Gustavo Deco
Center for Brain and Cognition
Universitat Pompeu Fabra, Barcelona
[email protected]