Abstract:Medical imaging denoising faces great challenges, yet is in great demand. With its distinctive characteristics, medical imaging denoising in the image domain requires innovative deep learning strategies. In this study, we propose a simple yet effective strategy, the content-noise complementary learning (CNCL) strategy, in which two deep learning predictors are used to learn the respective content and noise of the image dataset complementarily. A medical image denoising pipeline based on the CNCL strategy is presented, and is implemented as a generative adversarial network, where various representative networks (including U-Net, DnCNN, and SRDenseNet) are investigated as the predictors. The performance of these implemented models has been validated on medical imaging datasets including CT, MR, and PET. The results show that this strategy outperforms state-of-the-art denoising algorithms in terms of visual quality and quantitative metrics, and the strategy demonstrates a robust generalization capability. These findings validate that this simple yet effective strategy demonstrates promising potential for medical image denoising tasks, which could exert a clinical impact in the future. Code is available at: https://github.com/gengmufeng/CNCL-denoising.

What problem does this paper attempt to address?

### Problems the paper attempts to solve This paper attempts to solve the problem of medical image denoising. Noise in medical imaging can interfere with disease diagnosis and affect subsequent clinical decisions. Therefore, it is very important to remove this noise to restore high - quality medical images. Specifically, the paper proposes a simple and effective strategy - Content - Noise Complementary Learning (CNCL) strategy for medical image denoising. ### Background and motivation 1. **Importance of medical imaging**: - Computed tomography (CT), magnetic resonance imaging (MR) and positron emission tomography (PET) are three commonly used medical imaging modalities in clinical diagnosis. - These imaging techniques provide high - resolution structural information (CT and MR) and metabolic and functional information (PET). 2. **Noise problems**: - Low - dose CT examinations reduce ionizing radiation exposure by reducing the tube current or shortening the exposure time of the X - ray tube, but this will lead to a decline in image quality. - MR examinations may cause anxiety in patients due to the narrow space and long - term acquisition process, especially for patients with claustrophobia. - PET scans hope to reconstruct perfect images with a small number of coincidence events, but this also introduces noise problems. 3. **Limitations of existing methods**: - Traditional image processing algorithms (such as non - local means method, block - matching 3D method and diffusion filter) can reduce noise, but will lead to over - smoothing of the image and residual errors. - Deep learning methods show superior capabilities in medical image denoising tasks, but most methods either directly learn the content image or the noise image, and both have their own advantages and disadvantages. ### Proposed method 1. **Content - Noise Complementary Learning (CNCL) strategy**: - This strategy uses two deep - learning predictors to learn the content and noise of the image respectively, and performs denoising in a complementary manner. - By combining the advantages of content learning and noise learning, the CNCL strategy can better preserve the structure and contrast details of the image while stably removing noise. 2. **Implementation framework**: - Implement the CNCL strategy based on the generative adversarial network (GAN) framework, and regard medical image denoising as an end - to - end task. - Select three classic convolutional neural networks (U - Net, DnCNN and SRDenseNet) as predictors in the generator part, and comprehensively evaluate their denoising performance. ### Experimental verification 1. **Data set**: - Verify the effectiveness of the CNCL strategy on three medical image data sets (CT, PET and MR). - Also verify it on a natural image data set (smartphone image denoising data set, SIDD). 2. **Experimental results**: - The results show that the CNCL strategy is superior to existing denoising algorithms in both visual quality and quantitative indicators. - This strategy shows strong generalization ability in dealing with different types of noise. ### Summary This paper proposes a new Content - Noise Complementary Learning (CNCL) strategy. By combining the advantages of content learning and noise learning, it effectively solves the problem of medical image denoising. The experimental results verify the superior performance of this strategy on multiple medical image data sets, showing its potential in clinical applications.

Content-Noise Complementary Learning for Medical Image Denoising

CNN and Multi-Feature Extraction Based Denoising of CT Images

Multinoise-type Blind Denoising Using a Single Uniform Deep Convolutional Neural Network.

Frequency-Relevant Residual Learning for Multi-Modal Image Denoising.

Multi-Scale Detail–Noise Complementary Learning for Image Denoising

Cross-denoising Network against Corrupted Labels in Medical Image Segmentation with Domain Shift

A Novel Medical Image Denoising Method Based on Conditional Generative Adversarial Network

Multilevel Noise Contrastive Network for Few-Shot Image Denoising

Unsupervised low-dose CT denoising using bidirectional contrastive network

Medical image denoising using convolutional neural network: a residual learning approach

Neighboring Slice Noise2Noise: Self-Supervised Medical Image Denoising from Single Noisy Image Volume

A Novel Network for Low-Dose CT Denoising Based on Dual-Branch Structure and Multi-Scale Residual Attention

Continual medical image denoising based on triplet neural networks collaboration

Self-Supervised Joint Learning for pCLE Image Denoising

A blind medical image denoising method with noise generation network.

Self-supervised Noise2noise Method Utilizing Corrupted Images with a Modular Network for LDCT Denoising

Self-supervised deep learning for joint 3D low-dose PET/CT image denoising

Feature-Guided CNN for Denoising Images From Portable Ultrasound Devices

A Noise-level-aware Framework for PET Image Denoising

DCANet: Dual Convolutional Neural Network with Attention for Image Blind Denoising

Improving Medical Image Denoising via a Lightweight Plug-and-play Module