The impact of the parcellation strategy on the topological organization of brain networks in spinal cord injury.
1Kaushal M, 2Oni-Orisan A, 1Chen G, 1Kaushal W, 3Leschke J, 1Budde M, 4Schmit B, 1Li S, 1Muqeet V, 1Kurpad S
1Medical College of Wisconsin, Wauwatosa, WI, USA; 2UT Health Science Center at Houston, Houston, TX, USA; 3University of Minnesota, Minneapolis, MN, USA; 4Marquette University, Milwaukee, WI, USA
Objective: The impact of parcellation strategy adopted for the analysis of the resting-state functional MRI (rs-fMRI) datasets in evaluating functional connectivity alterations in the resting brain subsequent to spinal cord injury (SCI) remains unclear. The objective of the present study was to highlight the differences in the results derived for various network metrics with the use of two different parcellation schemes.
Design/Methods: After obtaining necessary IRB approval for the study, 15 subjects with chronic, complete cervical SCI and 15 neurologically intact controls were scanned with rs-fMRI. The raw data was preprocessed and parcellated according to the Automated Anatomical Labeling (AAL) atlas into 116 regions of interest (ROIs) or the functional atlas of Power et. al into 264 ROIs. Following the parcellation of the data, quantitative network metrics based on graph theory ware computed for evaluating the network attributes of integration and segregation. The network metrics of global efficiency (GE) and characteristic path length (CPL) were calculated for studying network integration and local efficiency (LE) and clustering coefficient (CC) were calculated for analyzing network segregation. These network metrics were computed for the whole brain network and compared between the SCI subjects and controls over a range of cost thresholds (0.1 to 0.2 with increments of 0.02).
Results: For the parcellation scheme based on the AAL atlas (116 ROIs), the metrics of GE and CPL used for examining network integration were not significantly different between the study groups. The metrics used for studying network segregation included CPL, which was not statistically significant and LE, which was only significantly reduced in the SCI group at cost threshold of 0.1 (p value < 0.05). For the parcellation based on the functional atlas (264 ROIs), both GE and CPL and LE were statistically significant between the study groups at all cost thresholds with GE showing an increase and CPL showing a decrease in the SCI group (p value < 0.05). The metrics of LE and CC were also found to be significantly reduced in the SCI group at all cost thresholds.
Conclusion: The results demonstrate that the choice of parcellation template impacts both the local and global topological properties of the human brain networks. The parcellation scheme of 264 ROIs showed significant differences for the all the network metrics between the two groups, which could be on account of more accurate segmentation of the brain as the ROIs in that atlas are derived from pooling of various studies. By highlighting the impact of the parcellation template on group differences, the study sheds light on the suitability of different atlases in analyzing attributes of brain networks in SCI. This could potentiate the analysis of the resting-state data by improving the interpretation of results with possible application in monitoring functional improvement conferred by various therapeutic approaches.
Support: Falk Foundation, Bryon Riesch Paralysis Foundation.
You Must Be Logged In
This Reminders tab will display for your scheduled lectures 15 minutes before start time and for new messages waiting.