As long as the IntendedFor field in the fieldmap JSONs are specified correctly, they will be associated only with the BOLD files they are intended for. You do not have to worry about unrelated fieldmaps being in the directory.
That is not possible via the command line, you would have to do that reorganization yourself after the fact.
This is not the normal behavior of FMRIPREP. The general behavior is to save the data in a specific FOLDER that you define in the command line. Inside FOLDER, the data is saved in the following form:
As you mentioned, the fid- in fmaps naming caused an error during the bids validation. So I had to add the flag of --skip-bids-validation to the FMRIPREP call.
Does the IndendedFor: option still work to find the desired fmap?
You could rename the files if you wanted to be BIDS-compatible.
Otherwise I think that it should work with --skip-bids-validation if that is the only error you get.
I had to change the naming of the fmaps as follows, since I didn’t want to use the --skip-bids-validation flag. As far as I understand it should also apply the field map correction by default if the data is in BIDs. The preprocessing finished without error, however in the logs I see that Susceptibility distortion correction: None. Could you please let me know how I should approach this to run the fieldmap correction as well.
How are you associated your fieldmaps to your BOLD files? Can you provide an example of a phasediff.json and the output of the tree command on this subject’s raw data directory?
I have applied the changes as you suggested. The four tasks (catloc, music, aliceloc and spatial) were preprocced. I am attaching screenshots of the SDC results of the two tasks, as you can see in the images there are artifacts. Could please let me know why this is happening?
This is the command I used:
I would like to report back regarding the recent unusual susceptibility distortion correction issue, with expanding brain to non-brain voxels.
I tried masking the magnitude images but there was no improvement in the results.
Based on the recommendation by @Marcel_Zwiers, here. I masked the phasediff image. I didn’t use the BIDScoin pipeline. I basically multiplied the phase-diff image with a masked-eroded magnitude image which I am also sharing with you. And the results look much better. I have used fMRIPrep version: 23.1.3.