Sonar improves sparse view CT scans with better structure and accuracy
SONAR: A Structure-Consistent Neural Operator for Null-Space-Aware Sparse View CT Reconstruction
Computer Vision and Pattern Recognition
Summary
Taking CT scans with fewer X-ray views helps reduce radiation but makes it hard to get clear images. The authors designed SONAR, a machine learning method that better fills in missing information while keeping the image’s structure accurate. Unlike past methods that struggle with errors and fixed formats, SONAR works across different settings and improves image quality significantly. It has been tested on both simulated and real clinical data, showing more reliable and sharper CT reconstructions.
What this means in practice
- •For medical imaging teams: Improve low-dose CT scans quality across different device settings by robustly reconstructing images from fewer X-ray views.
- •For medical device developers: Integrate SONAR into CT scanner software to enhance sparse-view image reconstruction, enabling faster scans with less radiation.$Commercial implications: SONAR enables a product that improves CT image reconstruction quality and speed, making it attractive for CT manufacturers and software providers.
Authors
Song Ni, Haijun Yu, Haodong Li, Changsheng Fang, Shuyi Fan, Yixing Huang, Hengyong Yu
Abstract
Sparse-view computed tomography (CT) reduces radiation dose and acquisition time but remains severely ill-posed because incomplete projections poorly constrain null-space information. Existing learning-based methods often estimate this information in high-dimensional image space, conflate physical measurement errors with prediction errors, and depend on fixed discretizations. We propose SONAR, a Structure-Consistent Neural Operator for Null-Space-Aware Reconstruction. Instead of recovering the full null-space component, SONAR predicts a low-dimensional null-space-aware representation from the acquired projections as pseudo-measurements. It separates measurement and pseudo-measurement residuals, lifts them into the image domain through physics operators, and applies independent neural operators to constrain their structural effects, thereby accommodating admissible errors while suppressing unsupported structures. To support cross-discretization reconstruction, an anisotropic U-shaped neural operator models the periodic angular and nonperiodic detector dimensions using direction-dependent continuous supports, while image-domain neural operators re-discretize continuous kernels on target grids. These components form an optimization-inspired unrolled network. Experiments on simulated AAPM and clinical MARS photon-counting CT data demonstrate consistent improvements across seen and unseen view settings and unseen image resolutions. On AAPM dataset, SONAR improves PSNR by 1.87~dB at 62 views and by 7.63~dB under zero-shot transfer to a $512\times512$ grid over the strongest competing methods. SONAR also achieves the best overall performance in all clinical settings evaluated, demonstrating accurate, structurally reliable, and discretization-robust sparse-view CT reconstruction.