PhyDiCT: Plug-and-Play CT Reconstruction from Sparse X-Rays via Differentiable Rendering and Strong Priors

2026-10-06

AI Summary

This study addresses the severely ill-posed inverse problem of 3D CT reconstruction from sparse X-ray projections by proposing a training-free generative reconstruction framework. The method constructs a differentiable physics-based forward model grounded in the Beer-Lambert law, integrates text-conditioned diffusion priors with Split Gibbs sampling, and incorporates a test-time refinement strategy. This design guides the generation of lung CT images consistent with observed measurements without requiring any fine-tuning. Experimental evaluations on public datasets demonstrate that the proposed framework significantly outperforms existing methods, achieving a 7.5% improvement in SSIM over fully trained approaches while effectively balancing high reconstruction quality with anatomical structural consistency.

Abstract

Reconstructing 3D Computed Tomography (CT) images from a few X-ray projections is a highly ill-posed inverse problem due to the loss of volumetric information. We propose PhyDiCT, a training-free framework that integrates a differentiable Physics-based forward model, grounded in the Beer-Lambert law, with a text-conditioned Diffusion as a strong prior to reconstruct 3D lung CT images. We refer to our approach as training-free since the prior model is used without fine-tuning, and our goal is to steer the denoising procedure to generate samples consistent with X-ray observations. We guide the diffusion generation using Split Gibbs sampling to jointly optimize for projection fidelity (reward) and consistency with prior knowledge. Also, we introduce a test-time refinement step that enhances image realism and anatomical coherence. We extensively evaluate our method on publicly available 3D CT datasets using both perceptual and semantic metrics, demonstrating that it surpasses existing plug-and-play diffusion and fully trained reconstruction approaches. Our findings highlight that combining a strong generative prior with the underlying physics of image formation substantially improves reconstruction quality, e.g., 7.5\% improvement on SSIM compared to full training methods. Code will be released at https://github.com/batmanlab/PhyDiCT.

Problem

• CT reconstruction
• sparse X-rays
• ill-posed inverse problem
• 3D imaging

Innovation

• Differentiable Rendering
• Diffusion Prior
• Training-free Reconstruction
• Split Gibbs Sampling
• Test-time Refinement