44. Dealing with noise in psychophysiological interaction analysis. Issue 3 (March 2015)

• Record Type: Journal Article
• Authors: Barton, M.; Marecek, R.; Rektor, I.; Mikl, M.
• Is Part Of: Clinical neurophysiology. Volume 126:Issue 3(2015:Mar.)
• Page Start: e45
• Page End: 229
• Publication Date: 2015-03
• Subjects: Neurophysiology -- Periodicals; Electroencephalography -- Periodicals; Electromyography -- Periodicals; Neurology -- Periodicals; 612.8
• DOI: 10.1016/j.clinph.2014.10.203
• Languages: English
• ISSNs: 1388-2457

Abstract

<abstract xml:lang="en" abstract-type="author" id="ab005"> <title> <x xml:space="preserve">Abstract</x> </title> <sec> <title id="st005">Background</title> <p id="sp005">In some fields of fMRI data analysis, it is apparent that a correct methodology is crucial to achieving meaningful results. This paper provides a first quantitative evaluation of the effects of different preprocessing and noise filtration strategies on the psychophysiological interactions (PPI) – method for analysis of fMRI data (<xref id="c0005" rid="b0005">Friston et al., 1997</xref>), where noise management is not yet established.</p> </sec> <sec> <title id="st010">Materials and methods</title> <p id="sp010">To assess these effects, both real fMRI data and simulated fMRI data were used. Two regions of interest (ROIs) were chosen for the PPI analysis on the basis of their engagement during the task. PPI terms were computed and used in a general linear model (GLM); group-level analyses followed. This first-level PPI analysis pipeline was performed for 32 different preprocessing and analysis settings, which included either data filtration with RETROICOR (<xref id="c0010" rid="b0010">Glover et al., 2000</xref>) or no such filtration; different filtration of the ROI "seed" signal with a nuisance data-driven time series; and the involvement of these data-driven time series in the subsequent PPI GLM analysis (<xref id="c0015" rid="b0015">Weissenbacher et al., 2009</xref>). The extent of the statistically significant results was quantified at the group level using simple descriptive statistics.</p> </sec> <sec> <title id="st015">Conclusion</title> <p id="sp015">We conclude that different approaches for dealing with noise in PPI analysis yield appreciably different results. We definitely recommend the usage of RETROICOR. Filtering the ROI signal with data-driven signals and adding these signals to the GLM for assessing the PPI effects is apparently influential, but it is not clear whether their usage improves results in all cases.</p> </sec> </abstract>