Parcellation of functional sub-regions from fMRI: A graph clustering based approach. (March 2019)

• Record Type: Journal Article
• Authors: Haq, Nandinee Fariah; Tan, Sun Nee; McKeown, Martin J.; Wang, Z. Jane
• Is Part Of: Biomedical signal processing and control. Volume 49(2019)
• Page Start: 181
• Page End: 191
• Publication Date: 2019-03
• Subjects: Functional MRI -- Brain connectivity -- Community detection -- Putamen; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2018.11.007
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: A novel, computationally efficient framework for parcellating a single brain region-of-interest into spatially-contiguous functional sub-regions is proposed. A connectivity network generation approach is proposed that takes into account the connectivity and spatial distance between voxels in the target ROI as well as their dissimilarity in connectivities with other brain reference ROIs. A community detection algorithm is applied to sub-divide the generated network into several functionally connected and spatially contiguous subROIs. The framework is applied to synthetic data where it shows favorable results in the presence of noise compared to current approaches. When applied to resting state fMRI data from nine healthy subjects, the putaminal region was parcellated into two functional sub-regions, consistent with prior anatomical knowledge. Abstract: We propose a framework for parcellating a single brain region-of-interest (ROI) into spatially-contiguous functional sub-regions (subROIs) – each consisting of one or more voxels – based on fMRI connectivity patterns between subROIs and other brain ROIs. First, a functional connectivity network between the voxels in the primary ROI is generated by taking into account the connectivity pattern within the primary ROI and all other ROIs, with a spatial constraint to ensure the spatial continuity of the final subROIs. A community detection algorithm is then applied to the associated adjacency matrix of the connectivity network to parcellate it into functional subROIs. As an illustrative example, the framework was applied to resting state fMRI data from nine healthy subjects to parcellate the putaminal region into two functional subROIs. Training on odd and even time points resulted in more than 98% concurrence of voxels assigned to the same cluster. The relative fraction of voxels assigned to each subROIs was also robust across subjects. As a general tool, the proposed framework has the potential to be integrated into studies investigating subROI alterations in neurological disorders.