Researchers Submit Patent Application, "Cbct Simulation For Ct-To-Cbct Registration And Cbct Segmentation", for Approval (USPTO 20240202943).

A patent application has been submitted for a computer-based method that generates simulated cone-beam computed tomography (CBCT) images using computed tomography (CT) images. The method involves converting the Hounsfield Units (HU) of a voxel in the CT image into attenuation coefficients, projecting the CT image to a projection image, adding artificial noise, and back-projecting the projection image to create a simulated CBCT image. This method aims to address the lack of CBCT data for training algorithms by using readily available CT data. The simulated CBCT images can be used for tasks such as CT-to-CBCT registration and AI-based segmentation. The patent application also includes claims for generating training data and registering CT images to CBCT images using artificial intelligence. [Extracted from the article]