Question about how to set parameters for multi-metric, multi-image registrations, and using the rigidity penalty term in itk-elastix

[tossin15]

I'm trying to replicate a pipeline (reference) that deformably registers two CT images of the abdominal aorta with their masks as follows:

1. Rigid registration of binary masks
2. "Implicit centerline alignment" - a Multi-metric, multi-image registration, where CT images are registered using mutual information, the masks are registered using bending energy, and a rigidity penalty is applied inside the mask
3. Deformation registration - Multi-metric, multi-image registration, where CT images are registered using mutual information, and the masks are registered using bending energy

I have looked at the elastix manual and the itk-elastix jupyter workbooks, but I'm still unclear about how the metrics are assigned. For instance, in the second step, my registration set up looks like this:

parameterObject = itk.ParameterObject.New()
pmap = parameterObject.GetDefaultParameterMap('bspline')

pmap['Registration'] = ('MultiMetricMultiResolutionRegistration',)
pmap['Metric'] = ('AdvancedMattesMutualInformation','TransformBendingEnergyPenalty','TransformRigidityPenalty')
pmap['Metric0Weight'] = ('1.0', )
pmap['Metric1Weight'] = ('10.0', )
pmap['Metric2Weight'] = ('20.0', )
pmap['NumberOfHistogramBins'] = ('32',)

pmap['FixedRigidityImageName'] = (fixedMaskFilePath,)
pmap['MovingRigidityImageName'] = (movingMaskFilePath,)

parameterObject.SetParameterMap(pmap)

elastixObj = itk.ElastixRegistrationMethod.New(fixedImage, movingImage)
elastixObj.SetParameterObject(parameterObject)
elastixObj.SetLogToConsole(True)
elastixObj.UpdateLargestPossibleRegion()

In this step, only the CT images have been added to ElastixRegistrationMethod as fixedImage and movingImage, so is it correct to assume that the mutual information and bending energy penalty terms will be applied to the CT images, while the rigidity penalty will be applied to the masks at the files pointed by FixedRigidityImageName and MovingRigidityImageName? Also, is this the correct/best way to apply to the rigidity penalty in itk-elastix? I wasn't sure if I could just use AddFixedImage() and AddMovingImage().

For step 3, my registration set up is as follows:

parameterObject = itk.ParameterObject.New()
pmap = parameterObject.GetDefaultParameterMap('bspline')

pmap['Registration'] = ('MultiMetricMultiResolutionRegistration',)
pmap['Metric'] = ('AdvancedMattesMutualInformation','TransformBendingEnergyPenalty')
pmap['Metric0Weight'] = ('1.0', )
pmap['Metric1Weight'] = ('100.0', )
#pmap['NumberOfHistogramBins'] = ('32',)

pmap['FixedImagePyramid'] = ('FixedSmoothingImagePyramid','FixedSmoothingImagePyramid')
pmap['MovingImagePyramid'] = ('MovingSmoothingImagePyramid','MovingSmoothingImagePyramid')
pmap['Interpolator'] = ('BSplineInterpolator','BSplineInterpolator')
pmap['ImageSampler'] = ('RandomCoordinate','RandomCoordinate')

parameterObject.SetParameterMap(pmap)

elastixObj = itk.ElastixRegistrationMethod.New(fixedImage, movingImage)
elastixObj.AddFixedImage(fixedMask)
elastixObj.AddMovingImage(movingMask)
elastixObj.SetParameterObject(parameterObject)
elastixObj.SetLogToConsole(True)
elastixObj.UpdateLargestPossibleRegion()

Here, does the MI metric only apply to the first set of images (fixedImage, movingImage) and the bending energy penalty only apply to the second set of images (fixedMask, movingMask)? Or do both metrics apply to both sets of images?