Dear Chris,

We have recently updated our scripts to run the dcm2niix v20241211 version, and it does not seem to be working for our DTI scans anymore (Philips scanner, enhanced DICOMs). The output NIFTI is very dark and does not look as expected.

We were previously converting using v1.0.20170624, which worked great for the DTI scans (enhanced DICOM), but would like to be able to move to the latest converter to avoid any other issues cropping up (I am not an expert here so unsure if using a legacy version could cause other unforseen issues).

I have attached the output log from running dcm2niix v20241211 where you can see the warning messages about the DTI and the fieldmaps.

There are also warning messages for our SWI sequences so again, I'm not sure if these are problematic. Visually the NIFTIs look fine for SWIs...

On the upside, the v2024 seems to be converting our functional MRI sequences (where it seemed a bit problematic on the older version).

If you have any advice on what we could try or if we're missing something obvious, that would be much appreciated.

Thank you!

Best wishes

Cheryl

1. I would ensure you use the latest stable release (v1.0.20250506)
   
2. I do not have access to Philips hardware - if issues persist, feel free to share a sample dataset with my institutional email
     https://sc.edu/study/colleges_schools/ar...
   

Your images look fine. The issue is that Siemens DWI scans ALWAYS start with a b=0 volume, while this is not the case for Philips. Therefore, old versions of dcm2niix would reorder Philips volumes to make the first volume the b=0, which allows you to neglect the final parameter for FSL’s eddy_correct:
   eddy_correct <4dinput> <4doutput> <reference_no>

However, modern versions of FSL’s eddy/topup assume that data is provided in the order temporarlly acquired, modelling motion for artifacts. Due to this, modern versions of dcm2niix do not reorder data. Regardless of which verion of dcm2niix you use, the bval and bvec orders correspond to the image order, so things should process fine, albeit the new FSL tools might give a slightly more precise answer for newer data.

Note that with the old dcm2niix, the first volume is bright (as b=0 has higher SNR), while the subsequent volumes are dark. With the new version, the reverse is true, which you can see with the intensity timeline below.