[SimpleITK, Python, Registration]
I have a dataset of CT and CBCT scans; CBCTs don’t have a limited FOV, but they do contain a smaller portion of body compared to CT, i.e. they have less slices than CT scans.
What I want to do is to perform a rigid registration of a CBCT to a CT to see which part of the body in CT is also found in CBCT, and then get rid of the slices in CT that are outside of the “scope” of CBCT.
I don’t need the registered image as output at all, but only the original CBCT as well as the original CT without the slices that are out of the “scope” of CBCT.
Hope my explanation is clear enough. If anyone wonders why I’d like to do anything such – I’m trying to constrain the data that’s being seen by a generative adversarial network in order to prevent any bias when learning how to translate a CBCT to CT.
Since I would like it to be efficient, I thought I could ask here for some advice. Not very experienced with SimpleITK as I mostly used it as an interface for getting data in and out of my deep learning pipelines.
Maybe a way to use the data from
final_transform (a Transform class) from https://github.com/InsightSoftwareConsortium/SimpleITK-Notebooks/blob/master/Python/60_Registration_Introduction.ipynb (at the bottom) to figure out where along the z-axis is the relevant portion is located and truncate the original scan to it only?