Open TUYaYa1 opened 1 year ago
The proposed end-to-end registration framework uses the segmented CT bone to create projection DRRs. The reason we used segmentation rather than the whole volume is because we want to use the rigid bone features to perform 2D/3D registration. This is a common practice in conventional 2D/3D registration approaches.
This is not a requirement if you are just interested to project DRR images from an arbitrary CT volume. My take away of the reasons that your generated images look poor is because the field of view is outside of the CT volume (on the top right corner of the image suggests the boundary of your CT volume), so the contrast is very high. You may want to shift your volume to the center to get more realistic images.
Okay, I see. Thank you
Hi, When I use ProST to generate 2D images from CT, the quality of the generated images is poor. The CT I used is not segmented and contains other tissues except bones, but your source code uses the segmented image. I want to know whether the CT image must be segmented before it can be used? https://github.com/TUYaYa1/2D-3D/blob/main/CT-2D.png?raw=true