Cortical surface based identification of brain networks using high spatial resolution resting state fMRI data

Resting state fMRI (rsfMRI) has been demonstrated to be an effective modality by which to explore the functional networks of the human brain, as the low-frequency oscillations in rsfMRI time courses between spatially distant brain regions show the evidence of correlated activity patterns in the brain. This paper proposes a novel surface-based data-driven framework to explore these networks through the use of high resolution rsfMRI data. Guided by DTI defined fiber pathways and constrained by the gray matter, we map the rsfMRI BOLD signals onto the cortical surface generated by DTI-based tissue segmentation. We then use a data-driven affinity propagation clustering algorithm to identify these functional networks. Our experimental results demonstrate that the framework has high reproducibility and that several networks are detected reliably among individual subjects. Furthermore, our results exhibit that functional networks are highly correlated with structural connections. Finally, our framework is able to reveal visual sub-networks, indicating its potential role in sub-network exploration.