Preprocessing and Analysis of Resting-State fMRI

Preprocessing and analysis of resting-state functional MRI data was carried out using the CONN toolbox [44 (link)] and SPM12 (https://www.fil.ion.ucl.ac.uk/spm/). Preprocessing was performed according to CONN’s default preprocessing pipeline. Following coregistration, T1-weighted structural images were segmented and normalized to a template in Montreal Neurological Institute (MNI) space with a 1 x 1 x 1 mm resolution. T2*-weighted functional images were spatially realigned after coregistration and normalized to a MNI template with a 2 x 2 x 2 mm resolution using direct spatial segmentation and normalization. Functional scans were spatially smoothed using a 6 mm full-width half-maximum (FWHM) Gaussian filter. Outliers were defined using the artifact detection tool implemented in CONN. Here, outliers are identified using the observed BOLD signal and subject-motion in the scanner. Outliers were defined as a framewise displacement of at least 0.9 mm or a change of the global BOLD signal above five standard deviations. The framewise displacement is calculated based on 140 x 180x 115mm bounding box around the brain and estimating the largest displacement among six control points. The global BOLD signal change is defined as a change in the average BOLD signal within SPM’s global-mean mask scaled to standard deviation units. Framewise displacement as well as the global BOLD signal are computed at each timepoint. To deal with outliers, a variable number of noise components (one for each of the identified outlier) is used as potential confounding effect to account for any influence of the previously identified outlier scans on the BOLD signal [45 ]. Further denoising steps were applied to functional scans, including high-pass temporal filtering and linear detrending, removal of physiological confounds using the aCompCor method [46 (link)]. We also performed mainly two quality assurance checks to judge reliability of the imaging data. First, we checked, if the normalization process of structural and functional data was successful by generating figures of the normalized and realigned images (for each subject separately, but also averaged across all subjects). Secondly, we used quality assurance reports generated in CONN to examine the following parameters: the number of valid and invalid scans, the maximum, and mean extent of motion, but also the maximum and mean change of the global BOLD signal for every subject and the total sample.
Seed-based whole-brain correlations (Fisher-transformed bivariate Pearson correlation coefficient) were calculated in CONN using seed regions of interest (ROIs) from the Harvard-Oxford atlas. Based on the results of Maier et al. (2016), the following seeds were included in the analysis: Inferior frontal gyrus left and right, medial frontal gyrus left and right, superior frontal gyrus left and right, precentral gyrus left and right, insular cortex left and right, anterior and posterior cingulate cortex [7 (link)]. Significant associations between each ISAm score (total, hypokinesia and levodopa-induced dyskinesia) and functional connectivity between a respective seed ROI and every other voxel were evaluated using a general linear model in CONN. Age, gender, LEDD, and UPDRS-III scores were included as covariates in each analysis. All results of these multiple regression analyses were corrected for multiple comparisons using family-wise error (FWE) correction at cluster level.