The Influence of Preprocessing Steps on Graph Theory Measures Derived from Resting State fMRI

Abstract

Resting state functional MRI (rs-fMRI) is an imaging technique that allows the spontaneous activity of the brain to be measured. Measures of functional connectivity highly depend on the quality of the BOLD signal data processing. In this study, our aim was to study the influence of preprocessing steps and their order of application on small-world topology and their efficiency in resting state fMRI data analysis using graph theory. We applied the most standard preprocessing steps: slice-timing, realign, smoothing, filtering, and the tCompCor method. In particular, we were interested in how preprocessing can retain the small-world economic properties and how to maximize the local and global efficiency of a network while minimizing the cost. Tests that we conducted in 54 healthy subjects showed that the choice and ordering of preprocessing steps impacted the graph measures. We found that the csr (where we applied realignment, smoothing, and tCompCor as a final step) and the scr (where we applied realignment, tCompCor and smoothing as a final step) strategies had the highest mean values of global efficiency (eg). Furthermore, we found that the fscr strategy (where we applied realignment, tCompCor, smoothing, and filtering as a final step), had the highest mean local efficiency (el) values. These results confirm that the graph theory measures of functional connectivity depend on the ordering of the processing steps, with the best results being obtained using smoothing and tCompCor as the final steps for global efficiency with additional filtering for local efficiency.

Effects of spatial smoothing on group-level differences in functional brain networks

“…For example, Magalhães, Marques, Soares, Alves, and Sousa ( 2015 ) show that registration approaches have a strong impact on network formation, and Power, Barnes, Snyder, Schlaggar, and Petersen ( 2012 ) and Power et al ( 2014 ) highlight the importance of head motion correction since head motion can alter functional connectivity. Further, Murphy and Fox ( 2017 ) review the impacts of global signal regression in resting-state fMRI (rsfMRI), and Gargouri et al ( 2018 ) report that graph-theoretical measures of functional connectivity depend on the order and choice of preprocessing steps. Despite these findings, few researchers comment on the possible effects of their preprocessing choices.…”

Evaluation of different cerebrospinal fluid and white matter fMRI filtering strategies—Quantifying noise removal and neural signal preservation

“…Noise management is important especially in connectivity analyses, that is, when computing seed functional connectivity (FC) (Weissenbacher et al, ), in graph theory analysis based on FC (Gargouri et al, ) and in analyses studying task‐modulated intrinsic fluctuations where noise management also plays a role, for example, in psychophysiological interactions (Barton et al, ). This is because the complicated spatiotemporal structure of noise (Jo et al, ) introduces (even long‐distance) correlations to the data, and hence, the connectivity estimate can be biased.…”

Kernel machine tests of association between brain networks and phenotypes

“…The difference between simulation and real datasets could be due to a variety of factors, either working in isolation or compounded on one another. Several recently-published studies [7] [26] [27] have noted that choice of pre-processing pipeline can impact the results of an inferential analysis involving graph theoretic measures, especially in resting state fMRI. We have not studied the impact of different parameters within the same pre-processing pipeline nor the impact of an entirely different manner of pre-processing on the RPD.…”