Research on CBCT-to-CT Generation Based on Edge and Global Enhancement

This paper introduces a Pix2pix-based edge and global enhancement network (Pix2pix-EGE) for synthetic computed tomography (SCT) generation, which is designed to address the limitations of Cone-Beam CT (CBCT). Although CBCT offers a low radiation dose, it suffers from noise, blurring, and artifacts. By incorporating the strategy of edge supervision mechanism and global information extraction, Pix2pix-EGE addresses the issues of edge blurring and insufficient global contrast that exist in both the head-and-neck and pelvic regions. The network includes key modules such as the Edge Refinement Module (ERM) and Multi-Scale Global Context Enhancement Module (MSGCEM) to improve edge detection and global context information. Comparative experiments demonstrate that Pix2pixEGE outperforms existing models in both the head-and-neck and pelvic regions, achieving significant improvements in Mean Absolute Error (MAE) and Structural Similarity Index Measure (SSIM). The method offers a promising approach for precise medical image synthesis, potentially replacing traditional CT in radiation-based dose calculations and reducing patient exposure.