questions > dcm2nii DTI Data
Showing 1-3 of 3 posts
Feb 3, 2017 08:02 AM | Graham Warner - MGH
dcm2nii DTI Data
I'm working on converting a large set of DTI dicoms to nifti using
dcm2nii and while I able to successfully generate the .nii.gz files
for each set of dicoms I am only getting the .bval and .bvec files
for some of them. There doesn't seem to be a consistent vendor,
model, pulse sequence, diffusion direction, diffusion gradient, set
of b values, etc that predicts whether the .bval and .bvec files
will be generated so I was hoping I could get a list of what header
tags dcm2nii looks at to generate those files.
Thanks,
Graham
Thanks,
Graham
Feb 3, 2017 03:02 PM | Chris Rorden
RE: dcm2nii DTI Data
Graham-
It would help if you could provide a bit more information, e.g. "I am running dcm2nii 2May2016 on a Windows to convert data from our Siemens Trio with B19". However, I have a few thoughts:
1.) Have you tried dcm2niix (which comes with MRIcroGL) instead of dcm2nii (which comes with MRIcron)? While I hope my old software is mature and stable, vendors re-interpret the DICOM standard and therefore having modern software helps.
2.) The location of the DTI tags varies by vendor, for Siemens I peak inside the proprietary CSA header.
https://www.na-mic.org/Wiki/index.php/NAMIC_Wiki:DTI:DICOM_for_DWI_and_DTI
3.) My software only generates bvec/bval files when there is variability in the B-values. If you acquire a sequence with all B=0 values, no text file is created (as from a software point of view this looks like a fMRI sequence). If you acquire B=0 images in a separate session and combine them with b-weighted images for analyses you need to create bvec and bval files for your B=0 series (e.g. https://github.com/rordenlab/spmScripts/...). Most DTI sequences have at least one B=0 image as well as the B-weighted images, so it should detect normal DTI scans fine. I would recommend setting up your DTI sequence to have a few B=0 images embedded with the weighted images, as this will improve measures like MD and DK (for my hints on DTI see here http://www.mccauslandcenter.sc.edu/crnl/...).
I tested my software on sequences from all vendors
https://www.nitrc.org/plugins/mwiki/index.php/dcm2nii:MainPage#Diffusion_Tensor_Imaging
But it is possible that your vendor has introduced its own private tags or methods to encode diffusion data. If so you can either provide sample images or create an updated version of dcm2niix that fixes the problem and submit a pull request on Github.
-c
It would help if you could provide a bit more information, e.g. "I am running dcm2nii 2May2016 on a Windows to convert data from our Siemens Trio with B19". However, I have a few thoughts:
1.) Have you tried dcm2niix (which comes with MRIcroGL) instead of dcm2nii (which comes with MRIcron)? While I hope my old software is mature and stable, vendors re-interpret the DICOM standard and therefore having modern software helps.
2.) The location of the DTI tags varies by vendor, for Siemens I peak inside the proprietary CSA header.
https://www.na-mic.org/Wiki/index.php/NAMIC_Wiki:DTI:DICOM_for_DWI_and_DTI
3.) My software only generates bvec/bval files when there is variability in the B-values. If you acquire a sequence with all B=0 values, no text file is created (as from a software point of view this looks like a fMRI sequence). If you acquire B=0 images in a separate session and combine them with b-weighted images for analyses you need to create bvec and bval files for your B=0 series (e.g. https://github.com/rordenlab/spmScripts/...). Most DTI sequences have at least one B=0 image as well as the B-weighted images, so it should detect normal DTI scans fine. I would recommend setting up your DTI sequence to have a few B=0 images embedded with the weighted images, as this will improve measures like MD and DK (for my hints on DTI see here http://www.mccauslandcenter.sc.edu/crnl/...).
I tested my software on sequences from all vendors
https://www.nitrc.org/plugins/mwiki/index.php/dcm2nii:MainPage#Diffusion_Tensor_Imaging
But it is possible that your vendor has introduced its own private tags or methods to encode diffusion data. If so you can either provide sample images or create an updated version of dcm2niix that fixes the problem and submit a pull request on Github.
-c
Feb 21, 2017 02:02 PM | Chris Rorden
RE: dcm2nii DTI Data
Note my software only generates BVEC/BVAL when these are
meaningful. So you will get these files when you have a sequence
with varying b-value. Therefore, you will get these files for the
raw images acquired in a DTI sequence, but you will not get them
for images derived from your raw data (e.g. ADC, MD, FA, etc
volumes are a result of processing the raw data and these volumes
can not be described by a single bvec/bval as they are the result
of multiple values). I set my console up not to generate these
derived images, as you will derive better ones after running
TOPUP/EDDY on your images.
The one case where you might want to create BVAL/BVEC files is if you acquire a series with only B=0 and want to concatenate it with series that have both B=0 and B>0 images. It is often useful to acquire multiple B=0 images as averaging provides a more stable estimate for some measures (e.g. MD). Personally, I like to embed extra B=0 volumes into my DTI series
http://www.mccauslandcenter.sc.edu/crnl/tools/advanced-dti
but if you do acquire a series of exclusively B=0 images, you will need to create BVAL and BVEC files that are filled with zeros. I have a script that can do this for you
https://github.com/rordenlab/spmScripts/blob/master/nii_merge_dki.m
Again, I think this is appropriate behavior for my software, as these B=0 series do not have any B-weightings and generating BVAL files in this instance would mean you would also get these files when they are useless and distracting (e.g. fMRI, resting state).
The one case where you might want to create BVAL/BVEC files is if you acquire a series with only B=0 and want to concatenate it with series that have both B=0 and B>0 images. It is often useful to acquire multiple B=0 images as averaging provides a more stable estimate for some measures (e.g. MD). Personally, I like to embed extra B=0 volumes into my DTI series
http://www.mccauslandcenter.sc.edu/crnl/tools/advanced-dti
but if you do acquire a series of exclusively B=0 images, you will need to create BVAL and BVEC files that are filled with zeros. I have a script that can do this for you
https://github.com/rordenlab/spmScripts/blob/master/nii_merge_dki.m
Again, I think this is appropriate behavior for my software, as these B=0 series do not have any B-weightings and generating BVAL files in this instance would mean you would also get these files when they are useless and distracting (e.g. fMRI, resting state).