Good evening fMRI Experts,
I am conducting a series of physio-physiological interactions (PPIs) using resting state functional MRI (rsfMRI) data from the Human Connectome Project for Early Psychosis (HCP-EP). The data was collected in two sessions, with two acquisitions per session - one in the anterior-to-posterior (AP) phase-encoding (PE) direction, and one in the posterior-to-anterior ¶ PE direction.
The data has been converted to BIDS and preprocessed using fMRIprep.
Consistent with the approach for a typical first-level PPI analysis, I ran my models on each subject’s individual runs independently and averaged my interaction term betas across runs to get my per-subject PPI measures.
As I was checking the ROI overlays for my subjects’ BOLD data (which has been registered to MNI152NLin2009cAsym space) I noticed that the signal in the orbitofrontal cortex (OFC) on the PA-acquired image is much worse that that on the AP acquired images. Here is an example of what I mean:
This is concerning to me, because I am interested in the medial OFC (mOFC) as one of my ROIs. I am also interested in areas of the hindbrain as part of my analyses.
A qualitative inspection of the time-series information shows that the intensity values for the mOFC during PA scans is consistently lower than it is on the AP scans. Other ROIs I am interested are not effected in as consistent a manner insofar as they do not seem consistently higher or lower during one PE scan versus the other.
Furthermore, an inspection of the interaction betas across runs does not reveal a reliable difference between AP versus PA scans, though I have to testing this statistically yet.
My questions are as follows:
-
Is it still reasonable to average my interaction terms across runs given the difference in PE directions for the different runs? Note that I have other purely subcortical PPIs whose regions do not seem to suffer from the same signal disparities between PE acquisitions. I am wondering if the approach I’ve taken for those PPIs may still be valid? Is it “invalid” for the mOFC PPIs?
-
Is it reasonable to look at the mOFC at all, given how poor the signal is in the PA scans? Should I simply avoid it?
-
Would it be better if I took a hierarchical linear modeling approach to my analysis, where instead of collapsing across runs, I add the session number and PE direction as regressors in the model to account for session- and sequence-specific variance?
I look forward to your insights.
Kind regards,
Linda