Using DTI-TK & NODDI on multi-shell data

[Sila Genc]

Dear Gary and DTI-TK team,

I would like to know the best way of dealing with template construction and registration using multi-shell DWI data. 

The acquisitions I am working with have diffusion weightings of b1000 and b3000, with different TE/TR's and voxel size. I've normalised the b3000 to b1000 and run all of the NODDI processing steps without a glitch.

1. I am now creating a population-based template with my b1000 images, should I use the b3000 as well?

2. Is it easy enough to register the b3000 images directly to the population-based template? 

3. How can I register the ODI and FICVF maps from NODDI to the template?

Can this all be done with dti_diffeomorphic_population / dti_diffeomorphic_reg? If not which tool would you advise for the steps above?

Many thanks in advance,

Sila

[Hui Zhang]

Hi Sila,

This is a very good question and we have recently published a paper that has to consider all these aspects.  You can read it here:

http://link.springer.com/article/10.1007/s10545-014-9780-x

To answer your questions more generally:

1. I am now creating a population-based template with my b1000 images, should I use the b3000 as well?

This is what we have done.  I would avoid b=3000 because it is too noisy and missing a lot of GM contrast.

 

2. Is it easy enough to register the b3000 images directly to the population-based template?

This is definitely a no no.  It is far more accurate to register within subjects than between subjects.

What you have done seems reasonable given your constraints.  I'd normally recommend that you use the new FSL eddy tool (not to be confused with the old eddy_correct).  But this might be difficult in your case with different voxel sizes.

 

3. How can I register the ODI and FICVF maps from NODDI to the template?

Once you have aligned b=3000 to b=1000, the DTI volume and NODDI maps of each subject are spatially aligned and corresponding.  You can warp the NODDI maps to the template with the transformation you computed from registering the DTI volume to the template.

 

Can this all be done with dti_diffeomorphic_population / dti_diffeomorphic_reg? If not which tool would you advise for the steps above?

These do the job of aligning DTI volumes to the template.  See my answer above for the next steps.

Gary

[Sila Genc]

Hi Gary,

Thank you for your helpful advice. Unfortunately I am having issues with transforming the NODDI images to template space (step 3 above). I have tried using dti_warp_to_template, as well as deformationSymTensor3DVolume and get a 'Segmentation fault' error, which doesnt create the transformed "$i_diffeo.nii.gz" files. I've also tried deformationScalarVolume, however when I overlay the images they are in a completely different location and dont overlay on the template.

Do you think the error is because NODDI maps are scalar images? Is it perhaps having troubles because the "combined.df" transformations are tensor-based and not scalar? Or am I missing something? I know I can register NODDI maps to the template using ANTS, but I just wanted to keep it consistent with DTI-TK tools if there is a way.

Many thanks,

Sila

[Niall Bourke]

Hi Sila,

I am just wondering if you found the cause/solution to the images shifting to a completely different location when running deformationScalarVolume? 

As I am struggling to resolve the same issue now.

Best,

Niall 

[Sila Genc]

Hi Niall,

I believe I resolved it using the fsl command: fslorient -deleteorient $i.nii (apply it to your registration input & reference). 

If that doesn't work let me know and I'll have another look!

Warm regards,

Sila 

--

---
You received this message because you are subscribed to a topic in the Google Groups "DTI-TK" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/dtitk/B37h_QktFHk/unsubscribe.
To unsubscribe from this group and all its topics, send an email to dtitk+un...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Do Tromp]

Hi Gary,

Thanks for all your great work and continued support on DTI-TK. I am interested in the questions posted here earlier this year, and would like to clarify a couple of steps. I looked at the paper you reference, and although helpful did not answer all my questions with regard to how you got the whole NODDI image in normalized space.

Am I understanding correctly that you run a tensor normalization with the b=1000 data (using the tensor model); and after this step you apply the combined affine & diffeomorphoic warp to the scalars of the diffusion weighted (non-tensor) images? And then once those are normalized you can model an ODF?

Thank you for clarifying,

Also, feel free to join the diffusion-imaging forum (groups.google.com/forum/#!forum/diffusion-imaging), your input would be greatly helpful.

All the best,

~Do