Hello. I want to apply an image registration on two brain MR T1images (say MRT1 A and MRT1 B after the names of the patients A and B). Additionally, A and B have their diffusion MRI brain image which are fully aligned with their respective MRT1. Given the transformation that was used to align MRT1 B to MRT1 A, is it possible to use the same transformation to transforma a diffusion MRI of patient B so that it aligned to the diffusion MRI of patient A?
Of course. That approach is routinely used. And not just for MRIs. The usual way of aligning PET images is to perform registration on their corresponding CT images, and then use that transformation. PET images have low spatial resolution and usually even worse anatomical representability.
Thank you for the information. By the way, I have read several examples of ITK and I have noticed that in the examples of mutual-information-based metrics the pixel rype is set to unsigned char. Does that mean that the images should definitely have pixels that are unsigned char too or is it possible to use images with float-numbered pixels?
Mutual information does intensity binning. But it is an internal detail. Mutual information metrics work well with short type, and should work with float, despite that type being used less frequently.
Suppose that i want to align images with float pixel type using mutual information based metric. Should i declare in the code that the pixel type is also float? Also, should the declaration of the pixel type in the code correspond to the pixel type of the input images?
Yes and yes.
Thank you and thank you 
1 Like