X-Splat: Gaussian Splatting for 3D CBCT Generation from Single Panoramic Radiograph

2026-07-02

AI Summary

This work addresses the challenges of geometric blurring and anatomical inconsistency in reconstructing three-dimensional dental CBCT images from a single panoramic radiograph—a highly underdetermined inverse problem. To this end, it introduces Gaussian splatting for the first time in this domain, initializing learnable anisotropic Gaussian primitives along X-ray paths according to known imaging geometry and optimizing them via a feedforward network. The method integrates Beer–Lambert reprojection constraints, multi-view radiographic supervision, and a lightweight residual refinement module to incorporate anatomical priors. Remarkably, without requiring real paired training data, the model substantially outperforms NeRF- and GAN-based baselines, accurately recovering fine structures such as teeth, cortical bone boundaries, alveolar morphology, and even the mandibular canal—previously difficult to reconstruct. A novel geometry-aware segmentation metric is also proposed for evaluation.

Abstract

Generating a 3D dental volume from a single panoramic radiograph (PXR) could provide a low-radiation alternative to Cone-Beam Computed Tomography (CBCT), but the problem is highly underdetermined: panoramic acquisition integrates 3D attenuation along curved X-ray paths into a 2D image, leaving depth-resolved anatomy unobserved. Existing implicit and generative approaches often produce oversmoothed geometry or anatomically inconsistent hallucinations, lacking geometry-driven supervision and relying on smooth representations unable to precisely localize sharp anatomical boundaries. We propose X-Splat, the first Gaussian Splatting framework for generating CBCT-like 3D dental volumes from a single PXR. X-Splat uses the known panoramic acquisition geometry as a generation scaffold: learnable anisotropic Gaussian primitives are initialized along the X-ray paths that formed the input image and adjusted in a single feed-forward pass, constrained by Beer-Lambert reprojection and multi-view radiographic training supervision. A lightweight residual refiner adds dataset-level anatomical priors without overriding the geometry already resolved by the Gaussians. We train on synthetic PXR-CBCT pairs, enabling direct volumetric supervision without paired real scans. We further introduce segmentation-based geometry-aware metrics, providing the first evaluation of PXR-based generation over maxillofacial anatomy. X-Splat outperforms NeRF- and GAN-based baselines, recovering individual teeth, cortical boundaries, and alveolar structure, including the mandibular canal which prior methods fail to reconstruct. Code will be available at https://github.com/tomek1911/X-Splat

Problem

• 3D CBCT generation
• panoramic radiograph
• Gaussian Splatting
• dental imaging
• anatomical reconstruction

Innovation

• Gaussian Splatting
• 3D CBCT reconstruction
• panoramic radiograph
• geometry-aware generation
• anisotropic Gaussians