I’ve been collecting some multi-echo fMRI data and finding that the final three volumes drop off rapidly. For other reasons, I’ve been trying several different protocols with varying TRs and TEs, but in every acquisition the final 3 volumes drop off for every echo as shown in the image below.
Simply trimming the final three volumes seems to work for downstream processing, but this takes a long time to manually do, especially across all the different echoes. fmriprep has an option to discard the first few volumes because that is a common preprocessing practice which would save me from doing it manually, but the option doesn’t work for final n vols.
I’m wondering if anyone else has had this issue or has any idea of what’s going on.
For reference, here are the parameters we are using have varied from around TR 1-3 s, TE 20-150 ms, GRAPPA 2-3, multi-band accel 4-6, and explored some other things as well. All lead to the same thing. I used dcm2niix for converting dicoms to bids. also tried heudiconv, again both resulted in the same thing.
Are you sure you don’t have no-excitation volumes enabled in your protocol? If you do, then you should split those out of the run and use the noRF suffix for them.
And do you happen to know if the noRF suffix should be in derivatives (since it technically has been modified) or should it be alongside the other niftis in the bids/sub-*/func directory? For purposes of bids compliance.
Thanks again. I’ve been wondering about this for months.
