Shared-Structure 4D Spectral Gaussian Representation for Sparse-View Spectral CT Reconstruction

2026-08-17

AI Summary

This study addresses the challenges of angular undersampling and spectral coupling in sparse-view spectral CT reconstruction by proposing a shared-structure 4D spectral Gaussian representation. The method decouples spatial geometry from spectral attenuation to construct a continuous 4D representation that enables querying unobserved channels, while integrating GSC-Net for efficient structure-spectrum decomposition modeling. Experimental results demonstrate that the proposed model achieves state-of-the-art reconstruction performance, attaining a PSNR of 36.61 dB, an SSIM of 0.914, and an LPIPS of 0.194. These findings confirm the method’s effectiveness in achieving accurate reconstruction under sparse-view conditions, successfully overcoming the inherent limitations of conventional approaches in handling coupled spectral and geometric information.

Abstract

Sparse-view spectral computed tomography (CT) reconstructs energy-resolved attenuation volumes from limited projection views, requiring simultaneous handling of angular undersampling and spectral coupling. We propose a SharedStructure 4D Spectral Gaussian Representation (4D-SG) that learns shared Gaussian geometry from full spectrum structural projections and uses a Gaussian-wise Spectral Density Curve Network (GSC-Net) to predict Gaussian raw density transformations. This factorization separates shared spatial structure from spectral attenuation variation, avoids independent channel geometry optimization, and establishes a continuous 4D-SG representation from discrete spectral measurements for unobserved spectral channel queries. Experiments on six synthesized, simulated projection, and real projection datasets with 50 views demonstrate the best average performance. Compared with the strongest Gaussian baseline, 4D-SG improves PSNR from 35.56 dB to 36.61 dB, increases SSIM from 0.909 to 0.914, and reduces LPIPS from 0.208 to 0.194, demonstrating its effectiveness for sparse-view spectral CT reconstruction.

Problem

• Sparse-view spectral CT
• Angular undersampling
• Spectral coupling
• Energy-resolved attenuation reconstruction

Innovation

• Shared-Structure 4D Spectral Gaussian
• Sparse-View Spectral CT
• Gaussian-wise Spectral Density Curve Network
• Spatial-Spectral Factorization
• Continuous Spectral Representation