Dcmstack

[moloney]

Interface to the "dcmstack" software.

[satra]

could you add an example here along the following lines. this serves two purposes:

• gives users a starting point on the website
• allows a demo of how diffusion vectors from siemens files could be decoded

Example ------- >>> from nipype.interfaces.dcmstack import LookupMetaInputSpec >>> res = LookupMetaInputSpec(in_file='diffusion_with_meta.nii', meta_keys = {'bvecs': 'CsaImage.DIffusionGradientDIrection'}) 

also if you change meta_keys to a dict, that would allow a user to rename the internal name to a more sematically useful form

[moloney]

I like the idea of allowing the user to map the meta data to shorter more semantically meaningful names. I will work on this.

I will also work on an example, but I know very little about the details of DTI. In particular it seems you need to do something additional with the DIffusionGradientDIrection to get the "bvecs" output by dcm2nii. I don't want to make a misleading or incorrect example.

[satra]

here is how nibabel does it for siemens.

https://github.com/nipy/nibabel/blob/master/nibabel/nicom/dicomreaders.py

and @rfdougherty just posted the ge version here:

https://github.com/cni/nims/blob/master/nimsutil/dicomutil.py

i think we could try to include bob's code in nibabel and then use that bit to do it for both siemens and ge. hopefully nolan can provide some info about philips.

[moloney]

The code in nibabel.nicom for Siemens DTI is not using the DiffusionGradientDirection at all, but rather computing the "q_vector" from the "B_matrix".

It would probably make more sense to define an interface specifically for getting this information, rather than trying to bake it into the generic LookupMeta interface (although I am open to suggestions).

[satra]

indeed - perhaps bvecs was a bad choice as an example. but the mapping idea would then allow extracting information that could then be used to generate other pieces of info.