Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Zero-shot CT Super-Resolution using Diffusion-based 2D Projection Priors and Signed 3D Gaussians

Jeonghyun Noh, Hyun-Jic Oh, Won-Ki Jeong

TL;DR

A novel zero-shot 3D CT SR framework that integrates diffusion-based upsampled 2D projection priors into the 3D reconstruction process and demonstrates superior quantitative and qualitative performance on two public datasets, and expert evaluations present the framework's clinical potential at 4x.

Abstract

Computed tomography (CT) is important in clinical diagnosis, but acquiring high-resolution (HR) CT is constrained by radiation exposure risks. While deep learning-based super-resolution (SR) methods have shown promise for reconstructing HR CT from low-resolution (LR) inputs, supervised approaches require paired datasets that are often unavailable. Zero-shot methods address this limitation by operating on single LR inputs; however, they frequently fail to recover fine structural details due to limited LR information within individual volumes. To overcome these limitations, we propose a novel zero-shot 3D CT SR framework that integrates diffusion-based upsampled 2D projection priors into the 3D reconstruction process. Specifically, our framework consists of two stages: (1) LR CT projection SR, training a diffusion model on abundant X-ray data to upsample LR projections, thereby enhancing the scarce information inherent in the LR inputs. (2) 3D CT volume reconstruction, using 3D Gaussian splatting with our novel Negative Alpha Blending (NAB-GS), which models positive and negative Gaussian densities to learn signed residuals between diffusion-generated HR and upsampled LR projections. Our framework demonstrates superior quantitative and qualitative performance on two public datasets, and expert evaluations present the framework's clinical potential at 4x.

Zero-shot CT Super-Resolution using Diffusion-based 2D Projection Priors and Signed 3D Gaussians

TL;DR

A novel zero-shot 3D CT SR framework that integrates diffusion-based upsampled 2D projection priors into the 3D reconstruction process and demonstrates superior quantitative and qualitative performance on two public datasets, and expert evaluations present the framework's clinical potential at 4x.

Abstract

Computed tomography (CT) is important in clinical diagnosis, but acquiring high-resolution (HR) CT is constrained by radiation exposure risks. While deep learning-based super-resolution (SR) methods have shown promise for reconstructing HR CT from low-resolution (LR) inputs, supervised approaches require paired datasets that are often unavailable. Zero-shot methods address this limitation by operating on single LR inputs; however, they frequently fail to recover fine structural details due to limited LR information within individual volumes. To overcome these limitations, we propose a novel zero-shot 3D CT SR framework that integrates diffusion-based upsampled 2D projection priors into the 3D reconstruction process. Specifically, our framework consists of two stages: (1) LR CT projection SR, training a diffusion model on abundant X-ray data to upsample LR projections, thereby enhancing the scarce information inherent in the LR inputs. (2) 3D CT volume reconstruction, using 3D Gaussian splatting with our novel Negative Alpha Blending (NAB-GS), which models positive and negative Gaussian densities to learn signed residuals between diffusion-generated HR and upsampled LR projections. Our framework demonstrates superior quantitative and qualitative performance on two public datasets, and expert evaluations present the framework's clinical potential at 4x.

Paper Structure

This paper contains 10 sections, 5 equations, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Zero-Shot 3D CT SR Framework
  3. LR Projection SR using Diffusion Model
  4. SR using DDNM.
  5. 3D CT Reconstruction via NAB-GS
  6. Negative alpha blending.
  7. Experiments
  8. Setup
  9. Results and Evaluation
  10. Conclusion

Figures (3)

  • Figure 1: Overview of our framework. (a) LR projection SR using diffusion model: A pre-trained diffusion model with 2D X-ray data is employed within the DDNM to generate HR 2D CT projection images from LR counterparts. (b) 3D CT reconstruction via NAB-GS: Using both positive and negative density Gaussians, we model a signed residual field between diffusion-generated HR projections and LR counterparts. For HR volume generation, the learned residual field is added onto the upsampled LR volume.
  • Figure 2: Visual comparisons of 3D CT reconstruction results. (a) Ground truth (GT), (b) Cubic interpolation, (c) ArSSR wu2022arbitrary, (d) CuNeRF chen2023cunerf, and (e) Ours. The green box and zoom-in highlight regions where our method excels at reconstruction. The error map is computed by the L2 norm between the prediction and the ground truth.
  • Figure 3: Visual comparisons of 3D CT reconstruction across activation functions. (a) Ground truth, (b) Softplus, (c) ReLU, (d) Sine, and (e) Ours (PReLU). Ours effectively enhances structural details while suppressing grainy noise.