Hi Jan
While Stam is right that you can get very different results depending on whether you interpolate data or transforms of the data (FA, Tensor, etc.), We think it is best to interpolate the data.
You can see an example of this in Karla's NeuroImage 2011 paper (PMID 21473920) - Figure 12.
In that example, data is interpolated from a sub-millimeter resolution to 2 or 3.5 mm, so the differences are huge. In practice, registration-related interpolation might not make a big difference (this is indeed what we find with TBSS). However, TBSS-2 will interpolate the tensors and re-calculate FA, which gives results that are closer to interpolating the data.
Cheers,
Saad.
On 12 Jul 2011, at 11:32, Jan Schreiber wrote:
> Hi Stam,
>
> thank you very much for your answer.
>
> Cheers,
> Jan
>
> ----- Ursprüngliche Mail -----
>> Hi Jan,
>>
>> The only issue is interpolation. I consider interpolation a delicate
>> business, especially when it comes to quantities like diffusion
>> tensors that do not leave in Euclidean space. You can get very
>> different results depending on the interpolation scheme and on whether
>> you interpolate the data or the tensor eigenvalues. Therefore, it is
>> safer to do the processing in diffusion space and transform the
>> results in standard space.
>>
>> Regarding the bvecs correction have a look at the archives, it has
>> been discussed a lot. Basically, it does not harm to correct the bvecs
>> after eddy current correction, but I have hardly seen any major
>> differences when you do so.
>>
>> Cheers,
>> Stam
>>
>>
>>
>> On 12 Jul 2011, at 10:32, Jan Schreiber wrote:
>>
>>> Dear Experts,
>>>
>>> there was a statement in this forum that the maps of FA should be
>>> calculated in diffusion rather than in structural or standard space.
>>>
>>> Why is this better?
>>>
>>> If one combines the eddy currents correction with the transformation
>>> to structural space only one interpolation step is required.
>>>
>>> Does the interpolation to a higher resolution affect the fitting of
>>> the diffusion tensor?
>>>
>>> Would you recommend to correct the bvecs after eddy currents
>>> correction in either diffusion or standard space?
>>>
>>> Thanks,
>>> Jan
>>>
>
--
Saad Jbabdi
University of Oxford, FMRIB Centre
JR Hospital, Headington, OX3 9DU, UK
(+44)1865-222466 (fax 717)
www.fmrib.ox.ac.uk/~saad
|