I’m new to DTI preprocessing and I’m trying to make sure I correctly understand the connectivity outputs generated by QSIRecon. From what I gather, the 4S256Parcels and 4S356Parcels atlases included with QSIRecon are (1) the Schaefer 200- and 300-ROI cortical atlases combined with a set of additional subcortical atlases, and (2) are defined in MNI152NLin2009cAsym space. I have a few questions about how to work with these outputs:
My fMRI data was preprocessed with fMRIPrep and is in MNI152NLin6Asym space. If I want to integrate the fMRI connectivity with the DTI connectivity, do I need to transform the fMRI outputs from MNI152NLin6Asym to MNI152NLin2009cAsym?
For the .mat connectivity matrices produced by QSIRecon: since they represent pairwise connection strengths between atlas regions, is it acceptable to use 4S256Parcels or 4S356Parcels during preprocessing and then drop the non-Schaefer parcels afterward if I only want cortical Schaefer 200 or 300 ROIs for downstream analyses?
If my main goal is to generate the structural connectivity matrix, is mrtrix_multishell_msmt_ACT-hsvs an appropriate choice for the --recon-spec option in QSIRecon?
Ideally, you would rerun fmriprep with the MNI152NLin2009cAsym output space, since transforming the BOLD after the fact is an extra interpolation you’d want to avoid. But also, you could warp the atlas to MNI152NLin6Asym. Since atlas parcel masks are binary, you aren’t as concerned about data interpolation.
Sure, they should always be the last 56 rows/columns.
If you have multishell DWI data and FreeSurfer outputs, yes.
My fMRIPrep outputs are in MNI152NLin6Asym and I’m using a Schaefer atlas in the same space. The structural DTI connectivity matrix I have is ROI×ROI in a .mat file, and as far as I understand, this matrix is atlas-dependent but not MNI-space dependent.
Since the ROI indices correspond to the same Schaefer parcels (after removing the non-schaefer 56 rows/columns), is it correct that I do not need to warp anything? In other words, because the connectivity matrix only depends on the atlas labeling—and not the template space—the connectome should align conceptually with the fMRI time series without any spatial transformations.
If you already have an atlas in the same space as your BOLD, then you’re fine. There may be other applications down the line for which it could be helpful for BOLD and DWI to be in the same space (e.g., visualization), but if you’re just analyzing the connectivity matrices this is not vital.