questions > Extracting bvals/bvecs from dwi missing Siemens ASCII header "ie the ASCCONV text file'
Showing 1-3 of 3 posts
Display:
Results per page:
Aug 15, 2018  10:08 AM | Karl Zimmerman - Imperial College London
Extracting bvals/bvecs from dwi missing Siemens ASCII header "ie the ASCCONV text file'
I've been receiving data from an external site for a study, which uses a SIEMENS 3T scanner. I've been using dcm2niix (v1.0.20180622) to convert the dicoms to NIFTI prior to running them through an analysis pipeline, however in some cases, no bvals/bvecs are extracted. After digging through why this might be the case, and having tried using other converters (eg. mri_convert), it appears the 'faulty' scans are missing the Siemens ACII header (later confirmed when trying to use nibabel to look at the headers).

My question is therefore if it is still possible to extract the bvals and bvecs for these scans using different settings/builds of dcm2niix, or if this means that it is not possible, and we should try get the centre to stop anonymising the data the way that they are doing currently which is causing these errors.

Thanks.
Aug 15, 2018  01:08 PM | Chris Rorden
RE: Extracting bvals/bvecs from dwi missing Siemens ASCII header "ie the ASCCONV text file'
There is no vendor agnostic standard for reading gradient directions from DICOM images. For Siemens data we typically extract bvals/bvecs by reading the proprietary CSA header. These fields are populated both by the vendor supplied sequences, as well as all the research sequences I have seen (e.g. the CMRR sequences). I suspect that the issue is that you have run your data through an anonymization scheme that has removed the CSA headers, that contain crucial details for the BIDS header. I suggest you look at the providence of your images - the raw DICOMs from the scanner will contain these. You want to update your pipeline to preserve these details. For the record, dcm2niix reads the stable binary portion of the CSA header, rather than the ASCII portion. 

You may want to examine your DICOM header (you can do this with the "-v 2" option in dcm2niix). The one hope is that your data still has tags 0019;000C and 0019;000D intact - in theory this would allow you to reconstruct the bval/bvec data. However, I suspect the tool that removed the CSA header also removed these. Even if you have these, other critical information is in the CSA header (e.g. slice timing).
Aug 16, 2018  08:08 AM | Karl Zimmerman - Imperial College London
RE: Extracting bvals/bvecs from dwi missing Siemens ASCII header "ie the ASCCONV text file'
Hi Chris,

Thanks for your reply and explanation - unfortunately as you predicted the 0019 tags have all been removed alongside the CSA tags. I'll contact the centre regarding their anonymisation process which I assume has been changed recently as the most recent images sent over have been successfully producing bvals and bvecs, and contain the CSA headers.