Research map / 3D vision / 3D generation
DiffuX2CT: Diffusion Learning to Reconstruct CT Images from Biplanar X-Rays
Xuhui Liu, Zhi Qiao, Runkun Liu, Hong Li, Xiantong Zhen, Zhen Qian, Juan Zhang, Baochang Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace simple 2D-encoder conditioning with the implicit conditioning mechanism (tri-plane generator + implicit decoder) (ablation) | 3D conditional diffusion model · LumbarV | PSNR: 29.37 -> 34.32 | Improved |
| add geometry projection loss (ablation) | DiffuX2CT · LumbarV | PSNR: 33.62 -> 34.32 | Improved |
| add geometry projection loss (ablation) | DiffuX2CT · LumbarV | FID: 16.22 -> 13.23 | Improved |
| add view modulators to fuse uw and vw planes (ablation) | Tri-plane decoupling generator · LumbarV | PSNR: 32.09 -> 34.32 | Improved |
| add view modulators instead of a ResNet block for plane fusion (ablation) | Tri-plane decoupling generator · LumbarV | PSNR: 33.07 -> 34.32 | Improved |
| add learnable uv embedding with Gaussian perturbation (ablation) | Tri-plane decoupling generator · LumbarV | PSNR: 34.19 -> 34.32 | Improved |
| use diffusion with ICM instead of regression ICM-Reg for downstream segmentation | DiffuX2CT · CTPelvic1K 3D segmentation | average DSC: 47.59 -> 74.10 | Improved |
| use 3D diffusion model instead of regression and GAN baselines | DiffuX2CT · CTSpine1K | FID: 50.63 -> 8.90 | Improved |
| DiffuX2CT vs regression X-CTRSNet on LIDC-IDRI PSNR3D | DiffuX2CT · LIDC-IDRI | PSNR3D: 22.35 -> 21.15 | Hurt |
| DiffuX2CT vs X2CT-GAN on LumbarV PSNR | DiffuX2CT · LumbarV | PSNR: 34.27 -> 34.32 | No clear effect |
Closest papers by meaning
- DiffNR: Diffusion-Enhanced Neural Representation Optimization for Sparse-View 3D Tomographic Reconstruction (AAAI 2026 · similarity 0.89)
- DM4CT: Benchmarking Diffusion Models for Computed Tomography Reconstruction (ICLR 2026 · similarity 0.88)
- DiffusionBlend: Learning 3D Image Prior through Position-aware Diffusion Score Blending for 3D Computed Tomography Reconstruction (NeurIPS 2024 · similarity 0.88)
- C^2RV: Cross-Regional and Cross-View Learning for Sparse-View CBCT Reconstruction (CVPR 2024 · similarity 0.86)
- Learning Image Priors Through Patch-Based Diffusion Models for Solving Inverse Problems (NeurIPS 2024 · similarity 0.86)
- DIFIX3D+: Improving 3D Reconstructions with Single-Step Diffusion Models (CVPR 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
