Registration for liver donor CT and MR

Dear community,

I am a hepatobiliary surgeon using 3D Slicer for preoperative planning. Until now, I have been segmenting the portal vein, hepatic vein, and liver parenchyma from the delayed phase of a CT scan for liver resection.

Currently, I am preparing for living liver donor surgeries and would like to perform 3D reconstruction of the liver hilum using:

  • Arterial and portal phases of CT (same session)
  • T2-weighted MR image acquired on a different date (to delineate the biliary structures)

However, registration has been extremely challenging, likely due to differing origin points and voxel spacing between the CT and MRI scans. I am not necessarily looking for an automatic registration method — I don’t mind a time-consuming or manual process — I just need a reliable way to align the CT and MRI so they can exist on the same coordinate system and allow accurate reconstruction.

Could anyone share a recommended workflow or technique in 3D Slicer (or any module/tool) to achieve this?

Thank you very much in advance for your help.

Best regards,

I’m not sure what you’ve tried so far, but if the patient is in close to the same position in both the MR and CT scan then you should get a reasonable registration with something like these steps (images and files are from some data from the SampleData module for illustration).

  • In the Data module, right click one of the volumes, say the CT

  • Select “register this” in the context menu

  • then right click on the MR and in the registration menu pick rigid registration

  • This will take you to General Registration with the volumes pre-selected and a transform created

  • Pick one of the alignment options, probably Geometry Align if the livers are roughly in the model of the scans.

  • Click apply and in a few seconds it should

  • To check, go back to the Data module. Click on “eye” column for the CT so it will be displayed in all slice views
    image

  • Right click on the “eye” for the MR and pick “show in slice views as foreground”

This should show you the images aligned.

If they don’t line up well, one common issue is too few samples. You can go back to the General Registration and change the Percentage from the default to a larger number. Maybe just going to 0.02 is enough but even going to 1% or more will not add too much compute time and can really improve the results.

image

If the livers are very differently shaped in the two scans due to breathing or patient pose, you might try a nonlinear transform. The steps are basically the same.