questions > Reordering of diffusion volumes after conversion
Showing 1-2 of 2 posts
Display:
Results per page:
Nov 30, 2018  09:11 AM | Josh Robinson
Reordering of diffusion volumes after conversion
Hello,

I am using dcm2niix (version jun 2018) to convert par/rec diffusion weighted data. I notice that the ordering of the data is different after conversion. The raw data is ordered starting with diffusion weighted images, trace/adc images, then the b0. After conversion, the b0, and trace/adc images are moved to the first 3 volumes in the nifti. For some other, it moves only the b0 to the front but otherwise keeps the same ordering of the data. Any changes that are made appear to also be made in the bvecs/bvals, but shouldn't the command keep the volume order the same from rec to nifti?

I haven't come across messages on the forum describing this problem, so I am mostly wondering if this normal behavior for dcm2niix? And if it is 'safe' to continue with analysis.

Any guidance is much appreciated.

Josh
Nov 30, 2018  12:11 PM | Chris Rorden
RE: Reordering of diffusion volumes after conversion
Josh

Notes on how dcm2niix handles PAR/REC sequences are here. Note that the PAR header can index images in a different order than they are stored in the REC file. I suspect if you look at your PAR files carefully you will see that dcm2niix is reordering them as directed in the PAR header. However, I believe this order is arbitrary, and may reflect details like when a volume was completed by reconstructors that were running in parallel with no bearing on true acquisition order.

Sample datasets demonstrate that the order that images are stored to disk and described in the PAR header do not reflect actual acquisition order. Ideally, it would be great if we could save these images in the same order as they were acquired (see Jesper Andersson's comments for issue 201). However, this is currently impossible with Philips PAR/REC data.

If possible, I would recommend you export the data to DICOM rather than PAR/REC format. I have tried to handle all the Philips PAR/REC data I have seen, but I do not have Philips equipment and there might be corner cases I am not aware of. I would suggest you acquire validation data as described in the 'dedicated document' on the dcm2niix wiki.