Zero-shot Volumetric CT Super-Resolution using 3D Gaussian Splatting with Upsampled 2D X-ray Projection Priors

2025-08-20

AI Summary

To address the strong reliance on paired low-resolution (LR)–high-resolution (HR) data and the difficulty in recovering high-frequency anatomical details in zero-shot CT super-resolution, this paper proposes a novel 3D reconstruction framework that requires no LR–HR paired data. Methodologically: (1) A diffusion model generates high-resolution X-ray projections as anatomical priors; (2) A negative-alpha blending mechanism (NAB-GS) is introduced to enable negative-value representation within Gaussian densities, facilitating residual learning between LR inputs and generated projections; (3) Per-projection adaptive sampling combined with 3D Gaussian rasterization enables volumetric reconstruction. Evaluated on two public datasets, our method significantly outperforms existing zero-shot and weakly supervised approaches, achieving substantial improvements in both quantitative metrics (PSNR/SSIM) and visual fidelity. To the best of our knowledge, this is the first work to realize high-quality, interpretable zero-shot 3D CT super-resolution reconstruction.

Technology Category

• Computer Vision: 3D Computer Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Security and Privacy: Data transparency and provenance
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Computed tomography (CT) is widely used in clinical diagnosis, but acquiring high-resolution (HR) CT is limited by radiation exposure risks. Deep learning-based super-resolution (SR) methods have been studied to reconstruct HR from low-resolution (LR) inputs. While supervised SR approaches have shown promising results, they require large-scale paired LR-HR volume datasets that are often unavailable. In contrast, zero-shot methods alleviate the need for paired data by using only a single LR input, but typically struggle to recover fine anatomical details due to limited internal information. To overcome these, we propose a novel zero-shot 3D CT SR framework that leverages upsampled 2D X-ray projection priors generated by a diffusion model. Exploiting the abundance of HR 2D X-ray data, we train a diffusion model on large-scale 2D X-ray projection and introduce a per-projection adaptive sampling strategy. It selects the generative process for each projection, thus providing HR projections as strong external priors for 3D CT reconstruction. These projections serve as inputs to 3D Gaussian splatting for reconstructing a 3D CT volume. Furthermore, we propose negative alpha blending (NAB-GS) that allows negative values in Gaussian density representation. NAB-GS enables residual learning between LR and diffusion-based projections, thereby enhancing high-frequency structure reconstruction. Experiments on two datasets show that our method achieves superior quantitative and qualitative results for 3D CT SR.

Problem

• Reconstructing high-resolution CT volumes from low-resolution inputs
• Overcoming limited anatomical detail recovery in zero-shot super-resolution
• Leveraging external 2D X-ray priors for 3D CT reconstruction

Innovation

• Uses 3D Gaussian splatting with upsampled 2D priors
• Leverages diffusion model for HR projection generation
• Introduces negative alpha blending for residual learning