Abstract
This study proposes an automatic panoramic X-ray image synthesis method from head CT data based
on mandible segmentation using TotalSegmentator. The method consists of three steps: segmentation of
anatomical regions, three-dimensional head tilt correction, and panoramic projection simulation. The mandible
and dentition masks are used to estimate the occlusal plane, enabling robust geometric correction independent
of patient-specific variations. The panoramic projection is simulated using an elliptical dental arch and an
asteroid-shaped rotation center trajectory, followed by edge-enhancement filtering for image clarity.
Experiments using 324 clinical CT datasets with corresponding panoramic radiographs showed that the proposed
method successfully generated panoramic images even for cases with tooth loss or metal artifacts, where
conventional threshold-based methods failed. Although the SSIM value (0.620 at 4° downward correction) was
slightly lower than that of conventional methods, the proposed approach demonstrated stable reconstruction and
potential applicability for data augmentation in deep learning–based dental diagnosis.
