Virtual Brain Inference (VBI), a flexible and integrative toolkit for efficient probabilistic inference on...

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Network neuroscience utilizes whole-brain network modeling to understand brain function and cognition.
• Virtual brain modeling integrates computational brain dynamics with individual neuroimaging data.
• A need exists for automated tools to estimate large-scale control parameters for these models.

Purpose of the Study:

• Introduce Virtual Brain Inference (VBI), an open-source toolkit for Bayesian inference on virtual brain models.
• Address the gap in automated model inversion for large-scale brain network analysis.
• Facilitate biophysically interpretable inference from diverse neuroimaging data.

Main Methods:

• Developed VBI toolkit with fast simulations, feature extraction, data handling, and probabilistic machine learning.
• Employed Bayesian inference for parameter estimation in virtual brain models.
• Conducted in-silico testing to validate inference accuracy and reliability.

Main Results:

• Demonstrated the accuracy and reliability of VBI for common whole-brain network models and neuroimaging data.
• VBI enables efficient and interpretable inference from non-invasive and invasive recordings.
• The toolkit supports uncertainty quantification for hypothesis evaluation.

Conclusions:

• VBI provides a flexible and integrative solution for virtual brain model inference.
• The toolkit has the potential to advance hypothesis testing in network neuroscience.
• VBI can contribute to precision medicine by enhancing predictive models of brain function.