A Denoising Framework for Real-World Ultra-Low Dose Lung CT Images Based on an Image Purification Strategy

2025-10-08

AI Summary

Ultra-low-dose CT (uLDCT) reduces radiation exposure but introduces severe noise, artifacts, and structural misalignment with normal-dose CT (NDCT), degrading the performance of existing denoising methods. To address this, we propose an image purification strategy that, for the first time, generates structurally aligned uLDCT–NDCT paired samples from real clinical data. Building upon this, we design a Frequency-domain Flow Matching (FFM) model that explicitly models the noise distribution in the frequency domain while enforcing anatomical structure consistency via frequency-aware constraints. The resulting dataset and FFM framework significantly improve the denoising performance of multiple state-of-the-art models on real-world uLDCT scans, achieving superior structural fidelity—setting a new standard in clinical uLDCT denoising. Both the dataset and source code are publicly released.

Technology Category

• Machine Learning: Structured Learning
• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Web data integration and cleaning

Abstract

Ultra-low dose CT (uLDCT) significantly reduces radiation exposure but introduces severe noise and artifacts. It also leads to substantial spatial misalignment between uLDCT and normal dose CT (NDCT) image pairs. This poses challenges for directly applying existing denoising networks trained on synthetic noise or aligned data. To address this core challenge in uLDCT denoising, this paper proposes an innovative denoising framework based on an Image Purification (IP) strategy. First, we construct a real clinical uLDCT lung dataset. Then, we propose an Image Purification strategy that generates structurally aligned uLDCT-NDCT image pairs, providing a high-quality data foundation for network training. Building upon this, we propose a Frequency-domain Flow Matching (FFM) model, which works synergistically with the IP strategy to excellently preserve the anatomical structure integrity of denoised images. Experiments on the real clinical dataset demonstrate that our IP strategy significantly enhances the performance of multiple mainstream denoising models on the uLDCT task. Notably, our proposed FFM model combined with the IP strategy achieves state-of-the-art (SOTA) results in anatomical structure preservation. This study provides an effective solution to the data mismatch problem in real-world uLDCT denoising. Code and dataset are available at https://github.com/MonkeyDadLufy/flow-matching.

Problem

• Addressing severe noise in ultra-low dose CT images
• Solving spatial misalignment between low and normal dose CT pairs
• Developing denoising networks for real clinical CT data

Innovation

• Image Purification strategy aligns CT image pairs
• Frequency-domain Flow Matching model preserves anatomy
• Framework solves data mismatch in real CT denoising