We preprocessed the fMRI scans using SPM12 (https://www.fil.ion.ucl.ac.uk/spm). Images were realigned for motion correction, and linear detrended over each series. For each participant, scans were spatially normalized in two steps, first coregistering the mean fMRI to the T1 anatomical scan and then warping to the “VBM8” template in Montreal Neurological Institute (MNI) space based on the T1 “DARTEL” spatial normalization algorithm [55]. These steps resulted in all participants’ fMRI images being in the template space.

The five major gyri were parcellated from the average of the high-resolution T1-weighted scans: Three short (anterior) gyri and two long (posterior) gyri: anterior short gyrus (ASG), mid short gyrus (MSG), posterior short gyrus (PSG), anterior long gyrus (ALG), and posterior long gyrus (PLG). We included these regions as mask files in nifti format (S2 File). Two experienced research team members determined the parcellation based on manual tracing with reference to a brain atlas [56]. The regions were outlined in normalized space; although this approach is slightly less accurate than individual tracing, the resolution of the fMRI data (>50 mm3) relative to the anatomical scans (< 1 mm3) is such that any differences in accuracy would not be meaningful. Signal intensity changes over time were extracted from each voxel in each gyrus from the processed images. For each gyrus in each participant, a mean time trend over all voxels was then calculated. Time trends were converted to percent change relative to the mean of the 1-minute baseline period. For each participant, the signals from the four challenges were separated and averaged to create one single handgrip percent change time trend that was passed to the group level analysis. While this averaging could theoretically result in reduced sensitivity, in practice, the statistical approach we chose takes advantage of repeated measures, and could detect small effect sizes.

To assess posterior-anterior effects, signal intensity changes were calculated relative to those in the PLG. As discussed above, the importance of the anterior insula has been described in clinical and animal studies. Our previous work showed this anterior-specific role could be demonstrated by comparing the fMRI signal in the anterior vs poster-most insula (PLG), and we repeated this technique here [46]. At each time point, signal intensity changes within the PLG were subtracted from those in the ASG, MSG, PSG and ALG for each hemisphere so that direct comparisons between these regions and the PLG could be assessed. Lateralization was assessed by subtracting signal changes in each of the five left gyri from the corresponding gyri on the right side; for example, ASG laterality was calculated by subtracting the left ASG time trend from the right ASG time trend.

The resulting fMRI signals were assessed for within- and between-group differences using repeated measures ANOVA (RMANOVA). The analysis was implemented with SAS “proc mixed”, as described earlier [57, 58]. In brief, this approach assesses within-group changes and between-group differences over time, with each 2 sec time-point during and after the challenge assessed relative to baseline time-points. We applied the Tukey-Fisher criterion for multiple comparisons; that is, we assessed the overall model for significance (p ≤ 0.05), and then effects of interest (time, group by time), before considering individual time-points of difference. The latter tests are performed within the “proc mixed” procedure, as the output includes time-point tests of significance (hence no post-hoc tests were needed). We assessed the effects in combined male-female models with sex as a covariate and in sex-specific models.

The RMANOVA mixed model approach allows for continuous variables to be included, so we performed secondary analyses of age and resting HR. We created four models that included different age effects added to the main model (group + time + group x time):

Main + age: age effects independent of group over the entire protocol, independent of time;

Main + age + age x group: group-specific effects of age over the entire protocol, independent of time;

Main + age + age x time: age effects on handgrip responses, independent of group.

Main + age + age x group + age x time + age x group x time: age effects on between-group differences in handgrip responses.

We repeated these calculations for models with HR in place of age. For the purposes of this study, we only focused on the within- and between-OSA and control group responses in the different models. The age-by-time and HR-by-time measures are not independent of the main effects of interest, but the degree to which these secondary models affect the within and between-group p-values reflects potential associations between the clinical and fMRI measures.

Note: The content above has been extracted from a research article, so it may not display correctly.

Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.

We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.