Self-supervised denoising of raw tomography detector data for improved image reconstruction

2025-11-21

AI Summary

Ultrafast electron-beam X-ray CT suffers from extremely low signal-to-noise ratio (SNR) in raw projection data due to ultrashort exposure times, resulting in severe noise and artifacts in reconstructed images. To address this, we propose two self-supervised deep learning denoising methods—Noise2Same and Neighbor2Neighbor—operating directly on unpaired raw detector measurements, eliminating the need for clean ground-truth labels. Furthermore, we introduce a joint optimization framework that simultaneously refines projections and reconstructions, enabling end-to-end noise suppression and reconstruction quality enhancement. Experimental results demonstrate that our approach significantly improves projection-domain SNR and consistently outperforms conventional non-learning denoising and iterative reconstruction methods in image quality, particularly under low-dose conditions where it exhibits superior robustness. This work establishes a generalizable, self-supervised learning framework for high-fidelity, low-dose industrial non-destructive testing and biomedical imaging.

Technology Category

• Machine Learning: Unsupervised & Self-Supervised Learning
• Computer Vision: Learning & Optimization for CV
• Search and Optimization: Non-convex Optimization

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Ultrafast electron beam X-ray computed tomography produces noisy data due to short measurement times, causing reconstruction artifacts and limiting overall image quality. To counteract these issues, two self-supervised deep learning methods for denoising of raw detector data were investigated and compared against a non-learning based denoising method. We found that the application of the deep-learning-based methods was able to enhance signal-to-noise ratios in the detector data and also led to consistent improvements of the reconstructed images, outperforming the non-learning based method.

Problem

• Denoising raw tomography detector data to reduce noise
• Improving image reconstruction quality by reducing artifacts
• Comparing self-supervised deep learning with traditional denoising methods

Innovation

• Self-supervised deep learning denoises raw detector data
• Deep learning methods enhance signal-to-noise ratios
• Denoising improves reconstructed images beyond conventional methods