Gaussian Splatting
Gaussian splatting for CT represents the volume as a set of explicit 3D Gaussian primitives, each with a position, shape and density, and optimises them so that their rendered X-ray projections match the measurements. It is adapted from 3D Gaussian splatting for photographic novel-view synthesis, replacing the emission-and-occlusion model with an attenuation model suited to transmission imaging.
Compared with implicit neural representations, the explicit primitives rasterise quickly on the GPU, so per-scan reconstruction can take minutes rather than hours, and the representation concentrates detail where the object needs it. Its main use so far is sparse-view CT and CBCT, with extensions to 4D respiratory imaging and metal artifact reduction.
Key technical issues are making the splatting integration exact enough for tomography, handling pose errors, initialisation from FBP or learned priors, and scaling to large volumes. The papers here range from X-ray novel-view synthesis to exact and scalable tomographic formulations.
Browse all 26 Gaussian Splatting papers in the searchable directory
Start here
If you are new to this topic, these are good first reads, chosen by the maintainers.
- Radiative Gaussian Splatting for Efficient X-Ray Novel View Synthesis (ECCV, 2024 · code)
Radiative Gaussian splatting for X-ray novel-view synthesis, the first adaptation of 3DGS to X-ray imaging, with code. - R2-Gaussian: Rectifying Radiative Gaussian Splatting for Tomographic Reconstruction (NeurIPS, 2024 · code)
Rectifies the splatting so it is consistent for tomography, the first 3DGS framework for sparse-view reconstruction, with code. - Learning 3D Gaussians for Extremely Sparse-View Cone-Beam CT Reconstruction (MICCAI, 2024 · code)
Uses 3D Gaussians as a feature representation for extremely sparse-view CBCT, with code. - X2-Gaussian: 4D Radiative Gaussian Splatting for Continuous-Time Tomographic Reconstruction (ICCV, 2025 · code)
Extends radiative Gaussian splatting to continuous-time 4D CT, with code. - FaCT-GS: Fast and Scalable CT Reconstruction with Gaussian Splatting (ECCV, 2026 · code)
A fast, scalable Gaussian-splatting CT reconstruction aimed at large volumes, with code.
Papers
All 26 papers, newest first.
Browse all 26 Gaussian Splatting papers in the searchable directory
Datasets, benchmarks and toolkits
Entries from the directory whose tasks or tags match this topic.
- Finnish Inverse Problems Society (FIPS) datasets (official page) · Dataset — Projection only. A family of real measured sinograms (walnut, lotus root, apple, carrot, phantoms) with carefully documented acquisition geometry — the cleanest public source for testing reconstruction operators.
- TIGRE · Toolkit — GPU-accelerated (CUDA) toolbox for cone-beam and parallel-beam CT; FDK, SART, SIRT, CGLS and TV variants.
- tomosipo · Toolkit — Painless specification of 2D/3D projection geometries; wraps ASTRA and integrates with PyTorch.
- torch-radon · Toolkit — Differentiable Radon transform and filtered backprojection in PyTorch.
Guides
Guides that cover this topic.
Other topics
- Surveys & Reviews (24)
- Classical & Iterative Reconstruction (71)
- Sparse-View CT (157)
- Limited-Angle Tomography (46)
- Low-Dose CT Denoising (69)
- Metal Artifact Reduction (23)
- Spectral & Dual-Energy CT (19)
- Cone-Beam CT (77)
- Dynamic & 4D CT (24)
- Scatter Correction (11)
- Geometry Calibration & Motion Compensation (28)
- Deep Unrolling & Model-Based Learning (68)
- Self-Supervised & Untrained Methods (35)
- Diffusion & Generative Priors (68)
- Implicit Neural Representations (25)