Introduction

Network neuroscience analyzes the connections between regions of the brain to assess the community structure of those connections and the role of different regions in communication across those communities. The regions in these communities may be defined by one of many existing parcellation definitions, which can be highly variable in their resolution and topography. Previous research has shown that choice of parcellation can influence network measures (Messé, 2019). Here, we asked if parcellation resolution influenced the community affiliated with a given brain location, and whether this related to changes in the average hubness of the regions in each community.

Methods

To assess how parcellation resolution influenced community assignment, we defined a set of surface-based coordinates of interest (COIs) and analyzed how many unique communities each was affiliated with across resolutions. To identify our COIs, we projected spherical volumes centered on a set of point locations (Power et al, 2011) onto a surface space and manually selected the coordinate located most centrally within the projected area. For the assignment analysis, we utilized the Global-Local parcellation set (Schaefer et al, 2018) in ten different resolutions from 100 to 1000 regions, each of which has each been mapped to a set of seven communities. We identified the parcel containing each of our COIs and its affiliated community at each resolution, then calculated each COI’s number of unique communities across resolutions. To assess how resolution influenced hubness measures of these communities, we applied a 17-community structure (Yeo et al, 2011) to the resting state data of 50 subjects from the Human Connectome Project and calculated coordinate-level metrics of participation coefficient (PC) and within-module degree (WMD). We assessed the PC and WMD for each parcel at each resolution by averaging across the values of all of the coordinates contained in that parcel. Finally, we assessed the trends in PC and WMD within each community across resolutions using repeated measures ANOVAs.

Results

Our network assignment analysis revealed that 161 (66.8%) out of the 241 COIs were affiliated with a single community across resolutions. Of the remaining 80 regions, 64 (79.6%) were affiliated with two communities; 15 (18.8%) with three; and one with four. Mapping out the affiliations between communities revealed a dense sharing structure. The somatomotor network (SMN) was the most isolated, sharing regions almost exclusively with the dorsal (DAN) and salience/ventral attention (SVAN) networks. The visual network primarily shared regions with the DAN, while the limbic system primarily shared with the default mode network (DMN). Higher-order association networks showed a rich pattern of mutual sharing. PC and WMD showed largely consistent values across resolutions in the control, limbic, DAN, SVAN, SMN, and visual networks. The DMN showed linear relationships for both WMD (F = 7.425, p = .009, η_p² = .134) and PC (F = 6.035, p = .018, η_p² = .112). That is, DMN appeared more internally integrated, but more externally isolated, with increasing parcel resolution.

Conclusions

The current results show that community affiliation is critically dependent on parcellation resolution while community-level network measures are not, which suggests that each parcellation is capturing different facets of a location’s connectivity based on which connections are included in the parcel definition. They also indicate that regions shift community affiliation along a constrained set of paths that are not described by aggregate measures of hubness. Together, this suggests that studies using a single parcellation to assign regions to unique communities do not accurately or fully capture the brain’s community structure, severely limiting their capacity to interpret regional roles within the network.

References

• Messé, A. (2019). Parcellation influence on the connectivity-based structure–function relationship in the human brain. Human Brain Mapping, (November), 1–14.
• Schaefer, A., Kong, R., Gordon, E. M., Laumann, T. O., Zuo, X.-N., Holmes, A. J., … Yeo, B. T. T. (2018). Local-Global Parcellation of the Human Cerebral Cortex from Intrinsic Functional Connectivity MRI. Cerebral Cortex, 28(9), 3095–3114.
• Yeo, B. T. T., Krienen, F. M., Sepulcre, J., Sabuncu, M. R., Lashkari, D., Hollinshead, M., … Buckner, R. L. (2011). The organization of the human cerebral cortex estimated by intrinsic functional connectivity. Journal of Neurophysiology, 106(3), 1125–1165.
• Power, J. D., Cohen, A. L., Nelson, S. M., Wig, G. S., Barnes, K. A., Church, J. A., … Petersen, S. E. (2011). Functional Network Organization of the Human Brain. Neuron, 72(4), 665–678.