BAI-Net: Individualized Anatomical Cerebral Cartography Using Graph Neural Network

Area of Science:

• Neuroscience
• Medical Imaging
• Computational Biology

Background:

• Brain atlases are crucial for neurological disorder diagnosis and treatment.
• Individual brain variations pose challenges for current clinical applications.
• Existing methods struggle to accurately map personalized brain regions.

Purpose of the Study:

• To develop an algorithm for subdividing the cerebral cortex into distinct areas based on individual brain morphology and connectomes.
• To create a reliable and explainable framework for individualized brain cartography.
• To improve the precision of neuromodulation and treatments by localizing them on individual brains.

Main Methods:

• The Brain Atlas Individualization Network (BAI-Net) algorithm was developed.
• BAI-Net integrates group priors from a population atlas.
• It adjusts areal probabilities using connectivity fingerprints from tractography and fiber-tract embedding.

Main Results:

• BAI-Net reliably individualizes brain areas across sessions and scanners.
• The algorithm outperforms conventional clustering methods in capturing heritable topographic variations.
• BAI-Net's topographic variability correlates with individual brain morphology, connectivity, cognition, and genetics.

Conclusions:

• BAI-Net offers an explainable framework for individualized brain cartography.
• This approach enhances the accuracy of localizing treatments and neuromodulation in individual brains.
• The findings suggest potential for more precise neurological therapies.