Registration about CT and CBCT

I have a set of CT images and CTCB images that need to be registered now, but they are not in the same physical space. How can I perform the conversion correctly? I know that CBCT has isocenter points, but CT does not. How can I place them in the same physical space? Thank you very much

Hello @ma1282029525,

To align the CBCT and CT images of the same object, standard rigid registration using correlation as a similarity metric should work (assuming the imaged object is rigid, e.g. the head and not a deformable one, e.g. the thorax). For more on registration, if you are working with ITK, please see the ITK book. If you are working with SimpleITK, please see the SimpleITK notebooks series 6*.

Before you write any code, I highly recommend trying out the registration options in Slicer.

1 Like

Maybe there is a problem with my description. We used ct machine and cbct machine to collect two sets of images respectively. The main parameters of the image collected by ct machine are origin -150 -150 -150, size 300 300 300 direction 100 010 001 and The main parameters collected by the cbct machine are -140 -140 -140 size 280 280 280 direction -100 0-10 001, which leads to them not being in the same physical space. What I want to ask is how to correct the two sets of machines to the same In space, I know that cbct has isocenter, but I don’t know how to convert the z-axis direction of ct

ITK performs all registration in physical space, so it does not matter that resolution of the images are different.

All you need for a successful registration is a good initial alignment (up to 5-10 mm translation, 5-10 degrees rotation is usually good enough) and you may need to crop the CT to approximately match the field of view of the CBCT.

The most robust way for initial alignment is landmark registration, based on 3-4 anatomical landmark points. Automatic methods based on geometric image center, image moments, etc. may work, too, but those depend on the image content.

1 Like

This is not what I want to ask. What I mean is that the physical position of my patient in the CT image is different from the physical position in the CBCT, and the difference is very, very large. I need to first move them to the same physical space, and then perform registration to use this registration data for treatment

Hello @ma1282029525,

You seem to be confused. Registration is the process that computes the transformation between the two coordinate systems. Classical registration usually has two steps, initialization and then nonlinear optimization. Please take a look at this Jupyter notebook from the SimpleITK tutorial and see if it clarifies things (I believe it will).

Now that I have figured out the problem, what I need to do is resample two images into the same physical space。