Fusing multi-scale functional connectivity patterns via Multi-Branch Vision Transformer (MB-ViT) for macaque brain...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Primate Research

Background:

• Brain age (BA) reflects brain maturity and is crucial for understanding development and neuropsychiatric disorders.
• Predicting BA in macaque models is vital due to their importance in biomedical research.
• Current methods for BA prediction in macaques are limited.

Purpose of the Study:

• To develop and validate a deep learning model for accurate macaque brain age prediction.
• To identify key brain regions and functional connections contributing to age prediction.
• To establish a foundation for studying age-related brain changes in primate models.

Main Methods:

• Utilized resting state functional magnetic resonance imaging (rs-fMRI) data from 450 rhesus macaques.
• Developed a novel Multi-Branch Vision Transformer (MB-ViT) model to fuse multi-scale brain functional connectivity (FC) patterns.
• Employed Gradient-weighted Class Activation Mapping (Grad-CAM) for identifying discriminative brain regions and connections.

Main Results:

• The MB-ViT model achieved high accuracy in predicting macaque brain age, with a correlation of 0.82 between predicted and chronological age (CA).
• Outperformed baseline models in terms of lower MAE and MSE, and higher PCC and R².
• Identified primary motor cortex (M1), visual cortex, posterior cingulate cortex (area v23), and dysgranular temporal pole as key predictive regions.

Conclusions:

• The MB-ViT model offers an effective and accurate method for predicting brain age in primates.
• This approach provides a valuable tool for future research into age-related neurological conditions in macaque models.
• The findings contribute to advancing our understanding of brain aging across species.