Progressive $\mathcal{J}$-Invariant Self-supervised Learning for Low-Dose CT Denoising

2026-01-20

AI Summary

This work addresses the challenge of denoising low-dose CT images in the absence of paired normal-dose data by proposing a self-supervised method that relies solely on low-dose scans. The approach introduces a progressive blind-spot denoising mechanism, which enforces conditional independence constraints to enable fine-grained noise modeling. To enhance generalization and mitigate overfitting, Gaussian noise is incorporated as a regularization strategy during training. Experimental results on the Mayo dataset demonstrate that the proposed method substantially outperforms existing self-supervised techniques and achieves performance comparable to, or even surpassing, that of several representative supervised denoising approaches.

Technology Category

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

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Self-supervised learning has been increasingly investigated for low-dose computed tomography (LDCT) image denoising, as it alleviates the dependence on paired normal-dose CT (NDCT) data, which are often difficult to collect. However, many existing self-supervised blind-spot denoising methods suffer from training inefficiencies and suboptimal performance due to restricted receptive fields. To mitigate this issue, we propose a novel Progressive $\mathcal{J}$-invariant Learning that maximizes the use of $\mathcal{J}$-invariant to enhance LDCT denoising performance. We introduce a step-wise blind-spot denoising mechanism that enforces conditional independence in a progressive manner, enabling more fine-grained learning for denoising. Furthermore, we explicitly inject a combination of controlled Gaussian and Poisson noise during training to regularize the denoising process and mitigate overfitting. Extensive experiments on the Mayo LDCT dataset demonstrate that the proposed method consistently outperforms existing self-supervised approaches and achieves performance comparable to, or better than, several representative supervised denoising methods.

Problem

• LDCT denoising
• self-supervised learning
• blind-spot denoising
• image denoising
• low-dose CT

Innovation

• self-supervised learning
• blind-spot denoising
• progressive denoising
• LDCT image denoising
• Gaussian noise regularization