Mapping effective connectivity by virtually perturbing a surrogate brain

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Effective connectivity (EC) is crucial for understanding brain function.
• Traditional EC methods are invasive or lack whole-brain coverage.
• A non-invasive method for whole-brain EC mapping is needed.

Purpose of the Study:

• To introduce Neural Perturbational Inference (NPI), a data-driven framework for whole-brain EC mapping.
• To develop a computational surrogate brain model for simulating neural dynamics.
• To enable causal inference of brain-wide interactions.

Main Methods:

• Developed an artificial neural network as a computational surrogate of the brain.
• Systematically perturbed regions in the surrogate model to analyze responses.
• Trained the network to model large-scale neural dynamics for EC inference.

Main Results:

• NPI demonstrated superiority over Granger causality and dynamic causal modeling in validation studies.
• Applied to resting-state fMRI data, NPI revealed consistent, structurally supported EC patterns.
• NPI-inferred EC closely matched real stimulation propagation patterns from evoked potential data.

Conclusions:

• NPI provides a powerful, non-invasive tool for mapping whole-brain effective connectivity.
• This framework facilitates a transition from correlational to causal understanding of brain function.
• NPI holds significant potential for neuroscience research and clinical applications.