BUG: Registration interface failed multi-modal

[hjmjohnson]

Registration interface was failing to support
multimodal input when they are given.

ex)
nipype:
antsWF.inputs.fixed_image=['fixed_t1.nii.gz',
'fixed_t2.nii.gz']
antsWF.inputs.moving_image=['moving_t1.nii.gz',
'moving_t2.nii.gz']

corresponding correct commandline:
--metric MI[
fixed_t1.nii.gz,
moving_t1.nii.gz,\ ...
--metric MI[
fixed_t2.nii.gz,
moving_t2.nii.gz,\ ...

Reported in issue: #1175

[hjmjohnson]

@satra Someone in my lab found a bug in using ANTs in a multi-modal registration. Is it OK to merge this bug fix?

[oesteban]

This looks OK to me. I was wondering, is it possible to use antsRegistration with multichannel images? Because, this is multimodal in the sense that you can create a composite metric that works in different modalities, right?

[satra]

@hjmjohnson - i'm not sure this is doing what you want it to do? it would also be good clarify how one uses it. i think the specs need to be clarified in the docstring and with examples.

for example, the old example1 makes no reference to fixed2 and moving2.

[oesteban]

@satra because it only uses the first modality, isn't it?

[satra]

correct - but that's why we need clear examples of how to do the multimodal stuff. also it seems example 5 should have been:
CC[fixed1, moving1] CC[fixed2.moving2] Mattes[fixed2. moving2]

[oesteban]

Well, @hjmjohnson will correct me, but that example would be monomodal operation... with a level-wise change of the images you try to register. I agree that this behavior is not clearly explained with antsRegistration and we have an opportunity here to make it clearer.