NITRC UNC-Wisconsin Neurodevelopment Rhesus MRI Database Forum: open-discussion

RE: phase encoding direction

Martin Styner — Mon, 13 May 2024 13:24:45 GMT

Reorientation of DWI data can indeed be an issue when bvecs are not correctly updated. I suggest you use one of the existing DWI tool datasets to process your data, as these tools know how to correctly adjust bvecs. Btw, if you have a re-orientation transform matrix and you apply it in Slicer via its resampling tool, the gradient vectors will get correctly updated if you have the DWI data as NRRD (I am not sure, but I doubt that Slicer is doing it correctly for nii/bvec/bval data).
Re eddy: Likely it does not matter, given you do not provide eddy with any information from topup.
Originally posted by tu ruilin:
<blockquote>
Hi, Martin
Thanks for the reply! As I am kind of new in DWI analysis, I have some quetions about these data, and apologize in advance for the following questions.
1. I noticed  the orientation of the raw DWI data differs from the processed data, so I reoriented the raw data using 3D Slicer and fslswapdim before preprocessing. However, I'm not sure if this method alters the bvecs. If so, is there a better way to reorient the data (the same orientation with the processed T1 images) and output the corresponding bvecs? 2. When using FSL's eddy for Eddy Current Correction, is it reasonable to set the parameter of acqp.txt to -1 0 0 0.1 (since there is no phase encoding direction, I arbitrarily chose one direction).
Best, Ruilin Tu
</blockquote>

RE: phase encoding direction

tu ruilin — Wed, 08 May 2024 15:39:45 GMT

Hi, Martin
Thanks for the reply! As I am kind of new in DWI analysis, I have some quetions about these data, and apologize in advance for the following questions.
1. I noticed  the orientation of the raw DWI data differs from the processed data, so I reoriented the raw data using 3D Slicer and fslswapdim before preprocessing. However, I'm not sure if this method alters the bvecs. If so, is there a better way to reorient the data (the same orientation with the processed T1 images) and output the corresponding bvecs? 2. When using FSL's eddy for Eddy Current Correction, is it reasonable to set the parameter of acqp.txt to -1 0 0 0.1 (since there is no phase encoding direction, I arbitrarily chose one direction).
Best, Ruilin Tu

RE: phase encoding direction

Martin Styner — Tue, 07 May 2024 3:35:27 GMT

Motion correction in dipy is probably fine (I have not used it before, so I have limited experience with it), you need to make sure it does also eddy current correction (which is commonly performed in the same step as motion correction via estimating an affine registration). We are either using our own implementation or the one in fsl's eddy. 
Martin

RE: phase encoding direction

tu ruilin — Mon, 06 May 2024 9:50:37 GMT

Hi, many thanks for your reply. if I don't use fsl or mrtrix3 to do eddy correction, but use DIPY for motion correction between volumes, is this okay?

RE: phase encoding direction

Martin Styner — Thu, 02 May 2024 19:37:05 GMT

The diffusion data was not acquired with a dual polarity encoding, but twice refocused acquisition, and thus there is only a single DWI acquired without polarity encoding. I.e., it does not have phase encoding directions that you can enter in mrtrix's preprocessing. You probably could try to enter any polarity encoding information. But the issue is that it will aim to correct distortions in that encoding direction, which might yield less than optimal results.

phase encoding direction

tu ruilin — Thu, 02 May 2024 7:19:27 GMT

Hello,
we want to estimate the orientation of the fibers in each voxel with constrained spherical deconvolution using Mrtrix. At the beginning, we would like to use files from AutoQc (e.g., *AutoQCed_DWI.nrrd), however, we could not convert  the NRRD data to Nifti using 3D Slicer.  We then choose to preprocess the origion data using mrtrix, which need to provide the phase encoding direction of these DWI datas. So would you please provide the phase encoding direction of these data for us. 
Thanks.

resting-state fmri data

tu ruilin — Thu, 25 Apr 2024 0:46:37 GMT

Hi, 
Many thank for your sharing, I was wondering if  rs-fmri data are available for these subjects?
Thanks,

RE: dMRI data in the PROCESSED folder

Roza Vlasova — Wed, 24 Feb 2021 21:11:09 GMT

Hi Brian,
I downloaded several DTI images from PROCESSED folder and checked them. They looks good (I attached screenshot from Slicer tool).
We don't use Diffusion Toolkit, but as I know from my experience with it - this tool supposed to reconstruct tensors from DWI images, so, you should provide preprocessed DWI images to have tensor image as output. Try to input to TrackVis DTI file directly, without Diffusion Toolkit step (because it's already tensors), if you will have problems with that, I suggest you to use files from AutoQC as input to Diffusion Toolkit. We have masks for structural data for the same subjects, you can register this masks to diffusion space.

Best,
Roza

Roza Vlasova PhD., Postdoctoral Fellow
Neuro Image Research and Analysis Lab
Departments of Psychiatry
University of North Carolina at Chapel Hill

dMRI data in the PROCESSED folder

Brian Jung — Wed, 24 Feb 2021 20:29:27 GMT

Hi, I have a question regarding the dMRI data in the PROCESSED folder.

I believe the data in the PROCESSED folder is an additional step from AutoQC. Specifically, I think it is a version of AutoQC data that has isotropic voxels. You can see this on the .nii and/or .nrrd metadata. Voxel dimensions in AutoQC are ~0.65 × ~0.65 × ~1.30 mm while in PROCESSED are ~0.65 × ~0.65 × ~0.65 mm.

The main reason we are interested in using the data in the PROCESSED folder is not because of the resolution, but because it already contains a brain mask (brain segmentation). Having this mask means we don't have to deal with fibers originating in surrounding (non-brain) tissues. That would also help for future co-registration to the structural MRIs.

Unfortunately, the DTI data (tensor data) in the PROCESSED folder seems to have problems. The tractography reconstruction generated in Diffusion Toolkit makes no sense: all fibers have the same orientation (see image attached). I initially thought this was a problem on my side (that I had used wrong parameters), but others get similar results.

So did the data in PROCESSED not undergo tensor reorientation during resampling and is actually not intended for analysis? Can the data in PROCESSED can be used for analysis and tractography reconstruction? Or are brain masks available for data in AutoQC?

DWI Orig data needs update

Martin Styner — Tue, 23 Feb 2021 22:49:14 GMT

Just a heads up to everyone: we need to update the original DWI data on the download page, as it is missing the raw data to go along with the header.