2.5. Data pre-processing and statistical analyses
Preprocessing and statistical analyses were conducted using SPM12 (http://www.l.ion.ucl.ac.uk/spm/software/spm12/). Functional data preprocessing included slice time correction, realignment to the first volume, and spatial normalization to the University of North Carolina at Chapel Hill neonate atlas (Shi et al., 2011). Motion artifacts were examined using the Artifact Detection Toolbox (ART) (https:// www.nitrc.org/projects/artifact_detect/). Volumes where global signal deviated more than two standard deviations from the mean signal or where the difference in motion between two neighboring volumes exceeded 1 mm were classified as outlier volumes. Subjects were excluded if the number of outliers in the fMRI data exceeded 30% of either rest or brushing blocks. The stimuli were modeled as one predictor convolved with the standard SPM12 hemodynamic response function. A fixed effects general linear model (GLM) analysis, including motion parameters and outlier volumes as regressors of no interest, was performed in each individual infant. The images of this first-level analysis were then used for the second-level group statistics in a new GLM. First, we ran a one-sample t test, controlling for infants’ gestational weeks and age at scan from birth, to test the effect of gentle stroking stimulation in the whole sample (N = 18). An a priori primary threshold for voxel-level statistical significance was set to p < 0.01 and results were FDR corrected at the cluster level (pFDR < 0.05), and a secondary threshold was set at p < 0.05, FDR corrected at the cluster level.
All the models were tested with the same thresholds also in a smaller group (N = 13), which is part of the whole sample (N = 18), in order to test the same models in slightly different samples and see if results would change. Results from the smaller sample (N = 13) are reported in Supplementary Material. Data from this smaller sample size was collected between the data included in the previous study conducted by Tuulari et al. (2019) and the final sample size (N = 18) presented here.