Unrecognized run structure from enhanced DICOMs

[helena-balabin]

General summary
Hi! I have been really enjoying bidscoin so far, I find it quite convenient 😄.

I am currently trying to convert some raw enhanced DICOMs to BIDS, but for some reason, bidscoin seems to struggle with recognizing the run structure in my functional MRI data.

The DICOM folder in the func/ subfolder of my raw data contains image files (IM_0001, IM_0005, IM_0008, ...) and some extra files (PS_0002, PS_0006, PS_0009, ..., XX_0003, XX_0004, XX_0007, ...) for a total of 12 runs of an identical fMRI sequence.

Currently, the respective func/ part in the resulting BIDS folder is looking something like this:

It seems that the runs are converted into suffices a-k instead of run numbers. Also, something else seems to be instead interpreted as the run number, because I end up with some files that contain "run-1" or "run-2" Interestingly, when I just run dcm2niix myself, I don't seem to have any issues:

I am not sure if this is expected behavior or a bug, but in any case, is there any chance I can replace the a-k suffices with appropriate run indices? Thanks a lot in advance!

Additional detail
I am running bidscoin 4.4.0 and dcm2niix v1.0.20241211 on Linux.

[marcelzwiers]

Hi there, hope I can keep you as a happy bidscoin user :-). In your output, there are a lot of files with a, b, c, etc suffixes. These are undesirable and are produced by dcm2niix if it sees that the output file already exists. This can happen either because you re-running bidscoiner without first cleaning up the bids output directory, i.e. without removing the output of the previous run. The other possibility is that multiple Series (images) are stored in the same folder, so dcm2niix has to add suffixes to the output name when it converts the data in the series folder (which is supposed to contain a single Series/image). So I need to know, do you still get the problem if you run bidscoiner without there being any output data?

[helena-balabin]

Thanks a lot for the quick reply 😄! I ran bidscoiner on an entirely new/empty output folder, so I don't believe it's related to any previous outputs. So if I understand correctly, it might instead be related to the fact that all my inputs (my 12 image files for the 12 runs) are stored in the same DICOM folder in the raw data?

[marcelzwiers]

Yes, your issue probably resides in bidscoin not properly understanding the structure of the input data. Can you show me how exactly your input data is organized (i.e. the tree of your root input folder)? And also what was the exact bidsmapper command that you used (in particular the sub/ses prefix options, if you used those)?

[helena-balabin]

My input (/raw) folder looks like this:

I did not use any specific sub/ses prefix options, I created an empty /bids output folder and ran the following:
bidsmapper raw bids
bidscoiner raw bids

[marcelzwiers]

I've never seen this structure used before, and it's fairly complicated with your multiple DICOMDIR files in there. It will probably help if you would take out your DICOM folders, and put the dcm files directly under func and anat. But even then, I can see that bidscoin may have trouble with the fact that you have multiple DICOMDIR files...

[marcelzwiers]

Perhaps try the above without the DICOMDIR files?

[marcelzwiers]

If I may ask, how did you get your data to be organized like that?

[helena-balabin]

I have tried removing the DICOMDIR files and putting the files directly under func/anat:

But I still run into the same issue:

This data structure is the one I got when exporting my MRI data from a Philips Achieva dstream 3T using the "enhanced DICOM" export option. I only added the subject, session, anat/func folder structure and exported the sequences to their respective subfolders. During my next scan, I will see what other options I might have to maybe end up with a less messy export format 😄

In the meantime, I will see if I can figure something out to get the run numbers right. Thanks a lot for all the help!

[marcelzwiers]

Did you get the data in ses-01 from a single export command or did you export the anat and func data separately? I'm asking because I assumed when exporting the data you always get a single DICOMDIR file...

[marcelzwiers]

You may have a look at the input structures that bidscoin understands and give dicomsort a try.

[helena-balabin]

I exported the anat and func data separately, maybe that's part of the issue. I'll try to group them, and/or see how I can use dicomsort to improve the structure.

[marcelzwiers]

Yes, that is confusing bidscoin. Perhaps you can try removing your DICOMDIR files and then run dicomsort...