Equivariance2Inverse: A Practical Self-Supervised CT Reconstruction Method Benchmarked on Real, Limited-Angle, and Blurred Data

2025/10/27 by Schut, Dirk Elias, Graas, Adriaan, van Liere, Robert +1
#FOS: Electrical engineering #Image and Video Processing (eess.IV) #electronic engineering #information engineering

paper · doi:10.48550/arxiv.2510.23317

Abstract

Deep learning has shown impressive results in reducing noise and artifacts in X-ray computed tomography (CT) reconstruction. Self-supervised CT reconstruction methods are especially appealing for real-world applications because they require no ground truth training examples. However, these methods involve a simplified X-ray physics model during training, which may make inaccurate assumptions, for example, about scintillator blurring, the scanning geometry, or the distribution of the noise. As a result, they can be less robust to real-world imaging circumstances. In this paper, we review the model assumptions of six recent self-supervised CT reconstruction methods. Moreover, we benchmark these methods on the real-world 2DeteCT dataset and on synthetic data with and without scintillator blurring and a limited-angle scanning geometry. The results of our benchmark show that methods that assume that the noise is pixel-wise independent do not perform well on data with scintillator blurring, and that assuming rotation invariance improves results on limited-angle reconstructions. Based on these findings, we combined successful concepts of the Robust Equivariant Imaging and Sparse2Inverse methods in a new self-supervised CT reconstruction method called Equivariance2Inverse.

Citations

• Equivariant Splitting: Self-supervised learning from incomplete data
• Noisier2Inverse: Self-Supervised Learning for Image Reconstruction with Correlated Noise
• Generalized Recorrupted-to-Recorrupted: Self-Supervised Learning Beyond Gaussian Noise
• UNSURE: self-supervised learning with Unknown Noise level and Stein's Unbiased Risk Estimate
• Perspective-Equivariance for Unsupervised Imaging with Camera Geometry
• Sparse2Inverse: Self-supervised inversion of sparse-view CT data
• X-ray Image Generation as a Method of Performance Prediction for Real-Time Inspection: a Case Study
• Diffusion Posterior Sampling for General Noisy Inverse Problems
• Improving Diffusion Models for Inverse Problems using Manifold Constraints
• A theoretical framework for self-supervised MR image reconstruction using sub-sampling via variable density Noisier2Noise
• Robust Equivariant Imaging: a fully unsupervised framework for learning to image from noisy and partial measurements
• Solving Inverse Problems in Medical Imaging with Score-Based Generative Models
• Equivariant Imaging: Learning Beyond the Range Space
• Deep Learning Techniques for Inverse Problems in Imaging
• Noise2Self: Blind Denoising by Self-Supervision
• Noise2Void - Learning Denoising from Single Noisy Images
• An apparatus and method for directly measuring the depth‐dependent gain and spatial resolution of turbid scintillators
• Deep Convolutional Neural Network for Inverse Problems in Imaging
• U-Net: Convolutional Networks for Biomedical Image Segmentation
• Adam: A Method for Stochastic Optimization
• Image quality assessment: from error visibility to structural similarity
• The inverse crime
• Estimation of the Mean of a Multivariate Normal Distribution
• The reconstruction of a three-dimensional structure from projections and its application to electron microscopy

Related

• Sparse2Inverse: Self-supervised inversion of sparse-view CT data
• A Self-Supervised Approach to Reconstruction in Sparse X-Ray Computed Tomography
• Equivariant Imaging: Learning Beyond the Range Space
• CNN-based regularisation for CT image reconstructions
• Noise2Filter: fast, self-supervised learning and real-time reconstruction for 3D Computed Tomography