Attention Mechanism in GAT Network Can Decode Age from fMRI Data

Brain networks differ in connectivity and function at different ages. In this research, we use functional Magnetic Resonance Imaging (fMRI) data to extract functional brain connectivity based on correlation. Our aim is to classifying brain networks based on their age in 3 classes using Graph Attention Network (GAT). Our results show that brain graphs are well classifiable in 3 distinct classes (3-5 years, 7-12 years, and adults) by GAT network with validation accuracy=72% and test accuracy=62.5%.