Panoramic X-Ray Synthesis From Dental CBCT Using Patient-Tailored Simulated Geometry With Uniform Projection Sampling

Panoramic X-ray (PX) is the most widely used imaging modality in dental health care. However, PXs can be severely degraded due to artifacts such as ghostly shadows, distortions, and superimposition of adjacent structures. Panoramic tomosynthesis from cone-beam computed tomography (CBCT) can be considered an alternative to noisy PX. Traditional cylinder-based methods take orthogonal projections along an appropriate dental arch; however, they suffer from a significant discrepancy between synthesized and genuine PXs, making them less applicable for clinical purposes. Recently, another approach based on simulated geometry that emulates the X-ray source-detector movement of PX scanners for ray tracing has been proposed. However, their methods suffer from poorly fitted ray trajectories for individual patients, uneven ray tracing, image misalignment, and, in some cases, the need for prior information on scan parameters. To tackle those problems, we propose a robust method of synthesizing PXs from CBCT by defining a universally applicable simulated geometry of the ray trajectory. Our method adjusts the ray trajectory to be optimal for a patient’s jaw shape and also achieves uniform ray sampling throughout the patient’s jaw by progressively increasing ray sampling intervals. In addition, tilt correction was applied to alleviate potential errors caused by incorrect patient positioning. Quantitative and qualitative results show that our method outperforms prior simulated geometry (SG)-based methods as well as image processing methods. Our framework can synthesize realistic PX images from dental CBCT by employing efficient ray tracing strategies customized to the patient’s oral structure without the need for scan parameters and ultimately empower dentists’ diagnostic capabilities.