New diffusion model synthesizes high-quality CT images from CBCT scans

Researchers have developed a novel diffusion-based conditional generative model, named EqDiff-CT, designed to synthesize high-quality computed tomography (CT) images from cone-beam computed tomography (CBCT) scans. This model utilizes a denoising diffusion probabilistic model (DDPM) with a group equivariant conditional U-Net backbone, incorporating e2cnn steerable layers to enforce rotational equivariance and cyclic C4 symmetry. Tested on the SynthRAD2025 dataset, EqDiff-CT demonstrated significant improvements in structural fidelity, Hounsfield Unit accuracy, and overall quantitative metrics compared to existing methods like CycleGAN and DDPM, leading to more realistic bone reconstructions and sharper soft tissue boundaries. AI

IMPACT This research could improve image quality in radiotherapy, potentially leading to more accurate dose calculations and adaptive planning in medical imaging.

RANK_REASON Academic paper detailing a new model and its performance on a specific task. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• Alessandro Perelli
• C4 model
• computed tomography
• cone beam computed tomography
• CycleGAN
• Denoising Diffusion Probabilistic Models
• e2cnn
• EqDiff-CT
• SynthRAD2025

• paper
• model release
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →