Predicting functional brain ROIs via fiber shape models.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Brain Connectivity Analysis

Background:

• Understanding human brain connectivity relies on accurately identifying Regions of Interest (ROIs).
• Current methods face challenges in precise ROI localization within individual brains.
• Multimodal neuroimaging data offers potential for improved ROI identification.

Purpose of the Study:

• To develop and validate a novel framework for predicting functional Regions of Interest (ROIs) in individual brains.
• To leverage multimodal task-based fMRI and Diffusion Tensor Imaging (DTI) data for enhanced ROI localization.
• To establish a robust method for analyzing structural and functional brain connectivities.

Main Methods:

• A novel ROI prediction framework was developed using learned fiber shape models.
• Training involved identifying ROIs as activation peaks in task-based fMRI data.
• White matter fiber shape models and ROI location distribution models were learned.
• ROI prediction was formulated as an energy minimization problem using learned models.
• Diffusion Tensor Imaging (DTI) data was utilized for ROI prediction in the individual brains.

Main Results:

• The framework achieved an average ROI prediction error of 3.45 mm.
• Performance was benchmarked against working memory task-based fMRI data.
• Promising results were demonstrated on the ADNI-2 longitudinal DTI dataset.
• The proposed method shows efficacy in localizing ROIs based on learned models.

Conclusions:

• The novel ROI prediction framework accurately localizes functional ROIs in individual brains.
• The integration of learned fiber shape and location distribution models enhances prediction accuracy.
• This approach offers a significant advancement for brain connectivity studies using multimodal neuroimaging data.
• The framework shows potential for clinical applications, as suggested by ADNI-2 dataset results.