Resolution-Independent Neural Operators for Multi-Rate Sparse-View CT

2025-12-13

AI Summary

Sparse-view CT reconstruction is severely ill-posed due to limited projection angles. Existing deep learning methods—such as CNNs and diffusion models—suffer from poor generalizability across sampling rates and image resolutions: CNNs rely on discrete convolution kernels incompatible with multi-resolution inputs, while diffusion models require fixed grids and incur high inference latency. This paper introduces CTO, the first neural operator specifically designed for CT, which pioneers continuous functional-space modeling for this task. CTO jointly learns the sinogram-to-image mapping via rotation-equivariant discrete-continuous convolutions, enabling truly resolution- and sampling-rate-agnostic reconstruction. Without retraining, CTO generalizes across diverse acquisition settings. In multi-rate and cross-resolution evaluations, it achieves >4 dB average PSNR gain over CNN-based methods and 3 dB higher PSNR with 500× faster inference than diffusion models, substantially enhancing clinical deployability.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Constraint Satisfaction and Optimization: Constraint Optimization
• Machine Learning: Feature Construction/Reformulation

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Security and Privacy: Data transparency and provenance

Abstract

Sparse-view Computed Tomography (CT) reconstructs images from a limited number of X-ray projections to reduce radiation and scanning time, which makes reconstruction an ill-posed inverse problem. Deep learning methods achieve high-fidelity reconstructions but often overfit to a fixed acquisition setup, failing to generalize across sampling rates and image resolutions. For example, convolutional neural networks (CNNs) use the same learned kernels across resolutions, leading to artifacts when data resolution changes. We propose Computed Tomography neural Operator (CTO), a unified CT reconstruction framework that extends to continuous function space, enabling generalization (without retraining) across sampling rates and image resolutions. CTO operates jointly in the sinogram and image domains through rotation-equivariant Discrete-Continuous convolutions parametrized in the function space, making it inherently resolution- and sampling-agnostic. Empirically, CTO enables consistent multi-sampling-rate and cross-resolution performance, with on average >4dB PSNR gain over CNNs. Compared to state-of-the-art diffusion methods, CTO is 500$ imes$ faster in inference time with on average 3dB gain. Empirical results also validate our design choices behind CTO's sinogram-space operator learning and rotation-equivariant convolution. Overall, CTO outperforms state-of-the-art baselines across sampling rates and resolutions, offering a scalable and generalizable solution that makes automated CT reconstruction more practical for deployment.

Problem

• Develops a resolution-independent neural operator for sparse-view CT reconstruction.
• Enables generalization across sampling rates and resolutions without retraining.
• Outperforms CNNs and diffusion models in speed and reconstruction quality.

Innovation

• Resolution-independent neural operators for CT reconstruction
• Rotation-equivariant discrete-continuous convolutions in function space
• Joint sinogram and image domain processing for generalization