An automated pipeline for constructing personalized virtual brains from multimodal neuroimaging data

Area of Science:

• Computational Neuroscience
• Neuroimaging Data Analysis
• Brain Modeling

Background:

• Vast amounts of multimodal neuroimaging data are generated annually.
• Standardized data fusion and reduction are crucial for extracting high-dimensional information for computational neuroscience.
• Existing methods require efficient tools for integrating diverse neuroimaging datasets into usable formats for brain modeling.

Purpose of the Study:

• To develop a standardized processing pipeline for multimodal neuroimaging data (MRI, EEG).
• To facilitate the creation of individualized large-scale brain network models using The Virtual Brain (TVB) platform.
• To address pitfalls in data processing and introduce novel methods for connectivity estimation.

Main Methods:

• Developed a processing pipeline integrating structural, functional, and diffusion-weighted MRI, and optionally EEG data.
• Utilized state-of-the-art neuroinformatics tools for parcellation, tessellation, and connectome generation.
• Implemented methods for estimating fiber tract transmission strengths and validating processing streams.

Main Results:

• The pipeline generates subject-specific parcellations, connectomes, and other essential data for TVB.
• Novel methods for estimating fiber tract transmission strengths were introduced and compared.
• Pipeline functionality was tested on 50 multimodal datasets, assessing rescan reliability.

Conclusions:

• The developed pipeline standardizes the construction of individualized brain models.
• The pipeline and processed data are publicly available, promoting reproducible research.
• The work provides principles for future standardization efforts in brain modeling.