Brain Age Prediction Using a Brain Multi-Atlas-Based Ensemble Deep Regression Method

This unpublished research article investigates brain age prediction from resting-state functional MRI using graph convolutional networks within a multi-atlas ensemble framework. The work compares population-based and individual-based graph representations, explores multiple parcellation schemes, and evaluates different feature representations and ensemble strategies. Using data from healthy subjects in the Human Connectome Project, the study demonstrates competitive brain age prediction performance, with the best model achieving a mean absolute error of 2.84 years. The manuscript highlights the potential of graph-based deep learning and multi-atlas modeling for functional neuroimaging-based brain age estimation.