Test-Retest Reliability of Functional Networks for Evaluation of Data-Driven Parcellation
- 677 Downloads
Brain parcellations play a key role in functional connectomics. A set of standard neuro-anatomical brain atlases are in common use in most studies. In addition, data-driven parcellations computed from fMRI data using a variety of clustering algorithms have also been used. Recent studies set out to determine the best parcellation in terms of quality and reliability have remained inconclusive without a clear winner. In this work, we investigated the utility of test-retest reliability of functional connectivity as an evaluation metric for comparing parcellations. Specifically, using data from the human connectome project, we compared a data-driven parcellation and a geometric parcellation using Intraclass Correlation Coefficient (ICC). We also investigated the impact of parcellation granularity on the test-retest reliability. We observed that the ICCs for geometric parcellation are better than those of a data-driven parcellation, suggesting that the FCs computed using regular parcels in the geometric atlases are more reliable than those computed using a data-driven parcellation.
KeywordsFunctional connectivity Test-Retest reliability Parcellation Precision neuroscience
This work was supported by NSF Grant IIS-1850204. The computational work is performed using the Data Analytics Cluster acquired through the Ohio Dept. of Higher Education’s RAPIDS grant in 2018.
- 1.HCP 1200 subjects data release. https://www.humanconnectome.org/study/hcp-young-adult/document/1200-subjects-data-release
- 2.HCP S1200 release reference manual. https://www.humanconnectome.org/storage/app/media/documentation/s1200/HCP_S1200_Release_Reference_Manual.pdf
- 3.Julia clustering package. https://github.com/JuliaStats/Clustering.jl
- 20.Meindl, T., et al.: Test-retest reproducibility of the default-mode network in healthy individuals. Hum. Brain Mapp. 31(2), 237–246 (2010)Google Scholar