help
60 Subscribers 
help > Getting erroneously low bvec values
Showing 1-2 of 2 posts
Aug 15, 2017 02:08 PM | Shereif Haykal
Getting erroneously low bvec values
Dear all,
I have a set DWI data in PAR/REC format. I used MRIcroGL (through the built-in dcm2niix) to convert my data to nifti format with accompanying bvec and bval files. After attempting to create FA maps from my nifti files, most of my data resulted in empty FA maps, showing basically the brain mask I used and no FA data.
After hours of trying to figure where the problem is, I realised that the nifti files which produced empty FA maps have abnormally low bvec values in their bvec files. All the nifti files which produced FA maps had bvec values somewhere between 0.09 and 0.9, while the ones which failed to produce FA maps had values around 0.001-0.05. I also noticed that the faulty bvec files had normal values in the third set of values (meaning after two "0" values).
In fact, when I used one of the "normal" bvec files to produce FA maps from the problematic nifti files, I got normal looking FA maps. So I'm almost certain the problem is with the bvec files.
This is very strange, as all the data was collected using the same scanner and protocol. Also, when I try to convert the PAR/REC files using the standalone dcm2niigui, I get no bvec/bval files at all!
Does anyone have any idea what I could be doing wrong? Any help is appreciated. I have attached an example of a faulty bvecs file to illustrate what I mean. Thanks in advance.
Best,
Shereif
I have a set DWI data in PAR/REC format. I used MRIcroGL (through the built-in dcm2niix) to convert my data to nifti format with accompanying bvec and bval files. After attempting to create FA maps from my nifti files, most of my data resulted in empty FA maps, showing basically the brain mask I used and no FA data.
After hours of trying to figure where the problem is, I realised that the nifti files which produced empty FA maps have abnormally low bvec values in their bvec files. All the nifti files which produced FA maps had bvec values somewhere between 0.09 and 0.9, while the ones which failed to produce FA maps had values around 0.001-0.05. I also noticed that the faulty bvec files had normal values in the third set of values (meaning after two "0" values).
In fact, when I used one of the "normal" bvec files to produce FA maps from the problematic nifti files, I got normal looking FA maps. So I'm almost certain the problem is with the bvec files.
This is very strange, as all the data was collected using the same scanner and protocol. Also, when I try to convert the PAR/REC files using the standalone dcm2niigui, I get no bvec/bval files at all!
Does anyone have any idea what I could be doing wrong? Any help is appreciated. I have attached an example of a faulty bvecs file to illustrate what I mean. Thanks in advance.
Best,
Shereif
May 8, 2018 12:05 AM | Chris Rorden
RE: Getting erroneously low bvec values
Hello-
Without seeing your files, it is hard to provide a clear solution. However, recently Philips have started to save images in the order they come off the reconstructor, rather than the order they were acquired or a spatially contiguous order. This required a new version of dcm2niix. Do also note that Philips have deprecated PAR/REC, so I would recommend saving Philips data in the enhanced DICOM format. Regardless, I would hope the latest pre-release of dcm2niix will read your old PAR/REC files fine, and if they don't, you should post an issue on the dcm2niix github web page.
You can get the pre-release of dcm2niix from here
https://github.com/rordenlab/dcm2niix/releases
I will make a new stable release once someone sends me a Philips DTI validation dataset - I do not have access to Philips hardware, so I am at the mercy of users to report issues and provide examples.
Without seeing your files, it is hard to provide a clear solution. However, recently Philips have started to save images in the order they come off the reconstructor, rather than the order they were acquired or a spatially contiguous order. This required a new version of dcm2niix. Do also note that Philips have deprecated PAR/REC, so I would recommend saving Philips data in the enhanced DICOM format. Regardless, I would hope the latest pre-release of dcm2niix will read your old PAR/REC files fine, and if they don't, you should post an issue on the dcm2niix github web page.
You can get the pre-release of dcm2niix from here
https://github.com/rordenlab/dcm2niix/releases
I will make a new stable release once someone sends me a Philips DTI validation dataset - I do not have access to Philips hardware, so I am at the mercy of users to report issues and provide examples.