EPC-3D-Diff: Equivariant Physics Consistent Conditional 3D Latent Diffusion for CBCT to CT Synthesis
The article presents EPC-3D-Diff, a new framework for synthesizing CT images from CBCT data. This method addresses issues related to scatter, noise, and artifacts in CBCT imaging, improving the accuracy of Hounsfield Units. The framework demonstrates significant performance enhancements over existing methods in both phantom and clinical datasets.
- ▪EPC-3D-Diff introduces a projection domain equivariance loss based on acquisition physics.
- ▪The framework utilizes a lightweight 3D autoencoder for efficient conditional diffusion in a compact latent space.
- ▪It achieved improvements of +7.4 dB in PSNR for phantom data and +1.8 dB for clinical data compared to state-of-the-art methods.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Computer Vision and Pattern Recognition arXiv:2605.20470 (cs) [Submitted on 19 May 2026] Title:EPC-3D-Diff: Equivariant Physics Consistent Conditional 3D Latent Diffusion for CBCT to CT Synthesis Authors:Alzahra Altalib, Chunhui Li, Haytham Al Ewaidat, Khaled Alawneh, Ahmad Qendel, Alessandro Perelli View a PDF of the paper titled EPC-3D-Diff: Equivariant Physics Consistent Conditional 3D Latent Diffusion for CBCT to CT Synthesis, by Alzahra Altalib and 5 other authors View PDF HTML (experimental) Abstract:Cone-beam CT (CBCT) is routinely acquired during radiotherapy for patient setup, but its quantitative reliability is degraded by scatter, noise, and reconstruction artifacts, limiting Hounsfield Unit (HU) accuracy.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
