fMRI Data to Connectivity Matrices: Building Functional Connectomes for Deep Learning Models

This study investigates to construct functional connectome of the human brain using fMRI data from healthy participants. The primary aim was to construct and analyze functional connectivity matrices to elucidate the intricate functional architecture of the brain. Participants performed a flanker task while fMRI data were acquired and subjected to a rigorous preprocessing pipeline. The preprocessed data were then used to create functional connectivity matrices using Pearson correlation coefficients to quantify temporal correlations between brain regions. The results revealed robust interconnections within the brain, with notable alignment to known functional networks, such as the visual, motor, and sensorimotor systems. Whole-brain connectivity mapping highlighted modular organization, identifying distinct functional modules, including the default mode, frontal-parietal, and sensorimotor networks. The study concluded that whole-brain functional connectivity mapping offers enhanced sensitivity in detecting subtle connectivity variations, underscoring the potential of integrating these connectomes with deep learning models, specifically Graph Neural Networks (GNNs), for advanced brain disease prediction. The use of GNNs can significantly improve the diagnosis and understanding of neurological and psychiatric disorders by capturing complex brain network relationships.