DKI analysis

[kamil....@gmail.com]

Dear All,

I'm struggling with DKI analysis. I obtained scans with bvals: 0, 1000 and 2000 s/mm2 in 64 dir with voxel size 2mm. 

After conversion from NIfTI to .mat I can't see FA maps. The output data is just black. After EPI/motion correction my data is... pink. 

I also tried to calculate tracts from raw data (black one) but ExploreDTI crushes while waitbar "One-off computation kurtosis metrics" appear.

Thank you in advance for your help.

Best regards,

Kamil Cepuch

[stijn m]

Dear Kamil,

Make sure you set ‘NaN’ as the b-value when converting your *.nii DWIs and *.txt B- matrix to a *.mat file. Convert the *.bval/bvec files to a b-matrix *.txt file first. After this step, the reordering is then automatically done for the b-matrix *.txt file. The B-matrix should also be correctly generated.

If looks like something went wrong in the first step of processing. Also make your the dicom to nifti conversion went fine.

Cheers,
Stijn

[kamil....@gmail.com]

Dear Stijn,
I followed steps you are writing about. I set 'NaN' bval and I generated b-matrix (I attached .bval, .bvec and generated .txt file). When it's about dcm2nii I used NiFtI files in FSL and they worked properly.

Best regards,

Kamil Cepuch

[Sjoerd Vos]

Hi Kamil,

One thing to check is how many b=0-images you tell ExploreDTI you have. Since you have multiple, 1 at the beginning and the rest at the end, it's easiest just to tell ExploreDTI you have 1. If you say you had 31 like you actually have (you probably do not need that many) then it might be confused by that in some of the steps... Or you can of course reorder the DWIs and b-matrix.

HTH,
Sjoerd

[kamil....@gmail.com]

Hello Sjoerd,

Indeed, reordering volumes by moving all b0 at the beginning of 4D file made my data work. But, unfortunately there is another problem with my files. I tried to export HARDI file to perform tractography and I get error "lmax should be in [0 -2] for this dataset" mentioned here. As my file is after MC I have no clue what can be done. I also tried to draw ROI and calculate tracts but again ExploreDTI crushes while waitbar "One-off computation kurtosis metrics" appear.

Best regards,

KC

[Sjoerd Vos]

Hi,

Unfortunately, ExploreDTI doesn't yet handle multi-shell data well for CSD. The CSD version is a single-shell method - as the original papers by Tournier et al. As a result, I think you'd have to make sure the mat-file contains only data with b=0 and bmax. The easiest way I think to do this is unfortunately not that easy:
- export stuff to nii (get b=0 + DWI and the b-matrix)
- Shuffle select 3D nifti volumes in 4D nifti (ensure you extract only the b=0 and bmax data and check that the corresponding b-matrix text-file is also created)
- convert that back into ExploreDTI mat-file to do CSD on.

Sjoerd

[kamil....@gmail.com]

Hello,

It's so bad to hear that. In this case have you got any recommendation which software con be suitable for that?  Beside ExploreDTI I'm starting to use FSL but as I'm a newbie I can swap to sth else.

Regards,

KC 

[Sjoerd Vos]

I think the current version of MRtrix also only allows single-shell data, but it's worth having a look on the MRtrix version 3 discussion forum. To be fully honest I personally prefer CSD over the FSL's bedpostx, as CSD appears to be more sensitive in detecting two and three-fiber crossings (e.g., Jeurissen et al, HBM 2013) so I'd suggest having a look at extracting single-shell data in ExploreDTI at least for a try on a single dataset.

Cheers,

Sjoerd

[kamil....@gmail.com]

Hey!
I tried to reconstruct tracts using CSD (with default settings) with my b-max scans from multi-shell acquisition. Unfortunately, it's not as perfect as I expected. There are many random fibers that shouldn't be there I guess.
When I perform DTI tracts seem to be ok. Can you give me any tips about crossing-fiber tractographys? 

Regards, 

KC