Integration of network topological features and graph Fourier transform for fMRI data analysis

Motivated by the recent progress in both graph signal processing and brain imaging, we integrate both techniques for complex brain network analysis. In particular, we address the challenge of evaluating the difference of functional connectivity networks between different age groups from resting state functional magnetic resonance imaging (RS-fMRI) observations. We proposed an approach to combine commonly used topological features from complex network analysis with the Graph Fourier Transform (GFT). Since GFT contributes to find the significant subspace of the original signal while topological features reveal the morphological structure of the brain network, they provide complementary information for characterizing brain networks. The method was validated on resting-state fMRI imaging data from Philadelphia Neurodevelopmental Cohort (PNC) dataset, comprised of normally developing adolescents from 8 to 22. The result shows that the model works well in distinguishing different age groups with an accuracy of 86.64% for 389 subjects.