I've completed a DTI analysis with FSL before (tutorial data) with no issues. However, when I use my labs data, FSL won't produce a proper output. Literally a blank white mask is shown when I enter DTI_FA into FSL View. When I enter the original unprocessed nii.gz image I see a normal brain scan.
I used this tutorial: http://www.cabiatl.com/CABI/resources/dti-analysis/
I downloaded the DCM files from my server (ep2d_DWI) scan and converted them to FSL_4D files with dcm2niigui giving me bvec and bval outputs as well.
I'm wondering if my issues are with the bvals, as it only has 3: '0,500,1000.' Where usually the tutorial files have 0, 1000 (repeated 20 or so times). In FSL View I see the following information when I only input the 4D file converted from DCM.
Voxels: 128x128x24
Dimensions: 1.71875x1.71875x6mm
Volumes: 3
Data type: signed short (16bpp)
Any Ideas??
Hi!
It is not clear to me if the problem occurs when you look at the DTI_FA.nii or at the Data.nii file.
DTI_FA is the output of dtifit which needs data.nii, bvecs and bvals as input. It has 1 volume.
Data.nii is the original data file which should have as much volumes as gradient directions (corresponding to the number of positions in bval and bvecs)
Your data description gives volumes = 3 which goes together with your bvals file (what is the bvecs file looking like?) wich however is only a simple diffusion volume with 3 directions, not enough to proceed. Also the dimention is highly anisotropic with 1.7 in plane and 6mm slice. I ma not sure if this is the real ep2d_DWI or not any other T2 file. Can you verify?
What does fslinfo DTI_FA.nii give you? Is the file empty or only badly scaled in FSLView?
Hope this helps,
Markus
Hi Shai,
It looks like there are only three images in the dataset, which is too few to estimate tensors. IIRC, dtifit will try to fit one anyway without error. If the b-vectors are collinear, that would explain the constant FA as well.
There could be more data to grab from the server, but it's more likely that it was a limited acquisition only meant to measure ADC.
Cheers,
Ryan
It seems that you downloaded DWI data for ADC map. You should use DTI data at least 6 gradient direction.
Markus: The DTI_FA.nii file only has 1 volume in it when I put it into FSL view..It's definitely a ep2d_DWI and not T2 file. There are 9 bvec values.
Ryan/Woo: I have an option to download two files from the server ep2d_DWI and ep2d_DWI_ADC - when I download the ADC file it doesn't give me a bvec/bval values after conversion.
Hi,
It seems that you have the wrong data. Your bval has 3 values, your bvec 9? They should have all the same as the number of volumes in the the converted FSL_4D file. Make sure you have the right data!
Best,M
The data was downloaded as DCM files directly from a folder of a single subject. Then I used dcm2niigui to convert them from dcm to nifti '.nii.gz' files which also gave me the bvec/bval values automatically. Is there another way to get the bvec/bval values? Could I just add 6 more "1000" values to the bval file?
I think the original DCM data you downloaded was not the dataset of DTI. The way of acquisition of DTI data by MRI is different from that of routine DWI image.
The problem is not the conversion of file type, but the DCM file you downloaded.
To understand if you are dealing with DWI or DTI data there are many ways.
A very basic one is to count the DICOM images in your B0 acquisition, the one similar to a conventional T2, and then check how many DICOM files are in your study.
if you are dealing with a DWI study the number of B0 images should be 1/3 of the total files (i.e. 50 out of 150).
If your are dealing with a DTI study the number of B0 images should be at least 1/6, or more, of the total files (i.e. 50 out of 300 or 50 out of 900 respectively for a 6 gradients or 18 gradients acquisition).
I like to use both MRTrix
https://github.com/jdtournier/mrtrix3
to look at new data, or rather to look at the info it contains with MRInfo, then to convert to whatever file type I need using MRConvert. You can use it to create a brain mask too, but we found the brain mask tool in FSL created a more continuous mask with less chance of islands appearing.
You could try using MRInfo to have a look at the file info and make sure you are looking at the file you think you are.
Indeed, dMRI ("DTI") data must have at least 6 directions to fit the matrix. My suggestion is as follows:
Download MRIcro. Then "Open", "Import", "Open foreign", then select any of your dicoms. Then, "Import", "Display/hide foreign header". Scroll down to the field entitled something like "Series description" - this field will tell you what modality the dicom is.
Make sure when conducting the FSL analyses that you are using the latest version (5.0.9). If using MRtrix3, make sure it is the latest version (0.3.15).
You can find further instructions in my handbook (http://jeromemallershandbookofstructuralbrainmrianalysis.yolasite.com/ ).
- Badges
- Science topic
- Similar topics
- Gels