Cone-beam Computed Tomography (CBCT) is widely used in dental imaging due to its relatively low radiation dose and cost-effectiveness compared to conventional CT. However, the limited field of view (FoV) often restricts scans to localized regions (e.g., a few teeth), which poses challenges for downstream tasks such as surgical planning and radiotherapy dose calculation. Moreover, CBCT images typically suffer from low soft-tissue contrast and artifacts, limiting their clinical utility. We address this dual challenge by proposing a 3D latent diffusion framework to extend truncated CBCT volumes toward full head-and-neck synthetic CT (sCT) volumes with enhanced image quality. Our volume compression network and latent diffusion model enable efficient volumetric synthesis with a reduced memory footprint and without requiring additional patient scanning. During inference, the diffusion model generates full-FoV sCT volumes conditioned on truncated CBCT inputs, providing a mechanism for FoV extension. Preliminary qualitative and quantitative results on the public SynthRAD2025 dataset demonstrate the feasibility of our approach in generating sCT images with extended FoV, highlighting the potential for future clinical applications despite current limitations.
