Pixelated DTI image

[Ehsan Misaghi]

Hi everybody,

I am almost new to DTI, so please excuse me for my question if it is too simple! I want to analyze two datasets, one of which has been taken using 1x1x5 voxel size and the other 2x2x2.5. I convert the DICOM files into NIFTI using MRIcron's dcm2nii utility and then convert the bvecs and bvals into bmatrix using ExploreDTI. Then I am converting NIFTI files into DTI .mat files using ExploreDTI. The 1x1x5 voxel size dataset is visually good when inputted into ExploreDTI, but unfortunately the 2x2x2.5 data is pixelated and it is almost in a non-usable state! I have attached an image from both of these datasets for your reference. I would be much obliged if anyone can say whether this is normal with this kind of data or there are any problems in the conversion pipeline I am using. BTW, I know that there may be some problems with the orientation in inputting the data. I think that this may not be the origin of the problem, even if it exists, but I might be wrong!

Thanks for your help in advance and sorry again if this is a simple question,

Ehsan

P.S: If it helps, the 1x1x5 dataset has 1 b=0 directions and 25 b=1000 directions with 27 slices each and the other dataset has 2 b=0 direction and 15 b=1000 directions with 53 slices each.

[stijn m]

Dear Ehsan,

Never hesitate to ask a question too simple. Your data looks fine, the pixelated image is due to the resolution. Best use an isotropic voxel instead of the huge difference in the 1x1x5 mm. For instance you might go for 2.5x2.5x2.5 and then register your DWI data to 1x1x1 anatomical T1 data. This is fairly common practice and ExploreDTI has great tools doing the job.
See Note number 10 in: Jones DK and Leemans A, "Diffusion Tensor Imaging", Methods in Molecular Biology 711:127-144, 2011.

Another consideration is to go for more directions, to have a better tensor estimate.

Cheers,
Stijn

[Ehsan Misaghi]

Dear Stijn,

Thank you for addressing my question. The solutions you spoke about are related to the imaging protocol itself, I think. Now that we have this dataset and we want to analyze it, anyway, what are your suggestions? I mean making the data as better as possible. Also, by note number 10 in the paper you mentioned, do you mean Eddy current/motion correction? If not, could you please let me know what exactly you mean?

Thanks,

Ehsan

[stijn m]

Dear Ehsan,

It's a bit of a trade off, but best go for the 2x2x2.5 considering that your voxels is more isotropic than the other. Best practice is to transform to T1 data, if you have that available. That way you should be able to make the best of it.
The reference to Jones DK and Leemans A, "Diffusion Tensor Imaging", Methods in Molecular Biology 711:127-144, 2011. not correct. Please check (with great thanks to the Providi Lab): http://providi-lab.org/onewebmedia/Jones11.pdf See Notes; number 10 at page 134. The whole article is a must-read considering your questions.

Cheers,
Stijn

[Ehsan Misaghi]

Hi Stijn,

Thanks a lot for taking the time to answer my question. I'll do the registration as you said. Sorry about the article, I had opened the wrong article! Anyway, thank you for all the information.

Regards,

Ehsan