[Insight-developers] NRRD Image Format (and tensor estimation thereafter?)

[Karthik Krishnan]

Done. You have the option of specifying the gradients in a a single 
VectorImage or as several itk::Image

One question:
Doesn't the he nrrd format for DWI images store the B0 and the 'n' 
gradient images in one nrrd image or are they in two seperate images ? 
Is that the NAMIC defacto ?

bye
karthik

Gordon Kindlmann wrote:

> hello,
>
>> Reference + DWI Gradient images  ---->  
>> DiffusionTensor3DReconstructionImageFilter  -----> itk::Image<  
>> DiffusionTensor3D< TScalar >, 3 >
>>
>> To this image you can apply one of the may tensor filters like:
>>
>> TensorFractionalAnisotropyImageFilter,
>> TensorRelativeAnisotropyImageFilter,
>> Hessian3DToVesselnessMeasureImageFilter,
>> SymmetricEigenAnalysisImageFilter,
>
>
> Vincent, is this helpful for your purposes, and do the different  
> types of image make sense?
>
>> I presume your question was "Can the DWI Gradient images" be a  
>> itk::VectorImage.
>> Not with the way the filter is written right now, It expects 'n'  
>> itk::Images as input, but its very easy to make it expect an  
>> itk::VectorImage as input. If you want, I can change that or add  
>> another filter.
>>
>> Would you rather have it expect a VectorImage ?
>
>
> Well, if its simply a matter of asking, then yes! :)
>
> I think it would be good to have the Tensor estimation filter take a  
> VectorImage, since this would allow one to write code that loads an  
> arbitrary DWI protocol, represented in the NAMIC Nrrd format, and  
> then estimate tensors and do other processing.
>
> Gordon
>
>
>