Functional parcellations affect the network measures in graph analysis of resting-state fMRI

There is a growing trend in the application of graph analysis to resting-state fMRI data. In such studies, vertices of the graph represent brain regions, and graph edges represent the connectivity between them. Regions are usually defined using anatomical atlases. In this paper we show that using functional parcellation which is considered to be better than anatomical segmentation causes differences in network measures of resting-state fMRI (rs-fMRI) graphs. In this study we used an anatomical atlas (AAL) and three functional parcellations with 98, 183, and 376 parcels for defining the brain regions in rs-fMRI data. Based on each, a functional connectivity graph is constructed and common network measures such as clustering coefficient, and characteristic-path length are calculated over 25 rs-fMRI data. Results indicate that networks obtained through functional parcellations have small world property at all resolutions. Correlation between network measures showed that characteristic path length in AAL-based network and parcellation-driven networks are noticeably different. This paper provides quantitative evidence on how using a functional parcellation, created from the functional data, can affect the measures that show the functional organization of the brain.