A dual-interactive fusion network for low-dose CT image denoising.

Low-dose computed tomography (LDCT) is an effective way to reduce radiation exposure for patients. However, it will increase the noise of reconstructed CT images and affect the precision of clinical diagnosis. The majority of the current deep learning-based denoising methods are built on convolutional neural networks (CNNs), which concentrate on local information and have little capacity for multiple structures modeling. Transformer structures are capable of computing each pixel's response on a global scale, but their extensive computation requirements prevent them from being widely used in medical image processing. To reduce the impact of LDCT scans on patients, this paper aims to develop an image post-processing method by combining CNN and Transformer structures. This method can obtain a high-quality images from LDCT. A hybrid CNN-Transformer (HCformer) codec network model is proposed for LDCT image denoising. A neighborhood feature enhancement (NEF) module is designed to introduce the local information into the Transformer's operation, and the representation of adjacent pixel information in the LDCT image denoising task is increased. The shifting window method is utilized to lower the computational complexity of the network model and overcome the problems that come with computing the MSA (Multi-head self-attention) process in a fixed window. Meanwhile, W/SW-MSA (Windows/Shifted window Multi-head self-attention) is alternately used in two layers of the Transformer to gain the information interaction between various Transformer layers. This approach can successfully decrease the Transformer's overall computational cost. The AAPM 2016 LDCT grand challenge dataset is employed for ablation and comparison experiments to demonstrate the viability of the proposed LDCT denoising method. Per the experimental findings, HCformer can increase the image quality metrics SSIM, HuRMSE and FSIM from 0.8017, 34.1898, and 0.6885 to 0.8507, 17.7213, and 0.7247, respectively. Additionally, the proposed HCformer algorithm will preserves image details while it reduces noise. In this paper, an HCformer structure is proposed based on deep learning and evaluated by using the AAPM LDCT dataset. Both the qualitative and quantitative comparison results confirm that the proposed HCformer outperforms other methods. The contribution of each component of the HCformer is also confirmed by the ablation experiments. HCformer can combine the advantages of CNN and Transformer, and it has great potential for LDCT image denoising and other tasks.

Positron emission tomography (PET) and computed tomography (CT) play a vital role in tumor-related medical diagnosis, assessment, and treatment planning. However, full-dose PET and CT pose the risk of excessive radiation exposure to patients, whereas low-dose images compromise image quality, impacting subsequent tumor recognition and diseasediagnosis. To solve such problems, we propose a Noise-Assisted Hybrid Attention Network (NAHANet) to reconstruct full-dose PET and CT images from low-dose PET (LDPET) and CT (LDCT) images to reduce patient radiation risks while ensuring the performance of subsequent tumor recognition. NAHANet contains two branches: the noise feature prediction branch (NFPB) and the cascaded reconstruction branch. Among them, NFPB providing noise features for the cascade reconstruction branch. The cascaded reconstruction branch comprises a shallow feature extraction module and a reconstruction module which contains a series of cascaded noise feature fusion blocks (NFFBs). Among these, the NFFB fuses the features extracted from low-dose images with the noise features obtained by NFPB to improve the feature extraction capability. To validate the effectiveness of the NAHANet method, we performed experiments using two public available datasets: the Ultra-low Dose PET Imaging Challenge dataset and Low Dose CT Grand Challengedataset. As a result, the proposed NAHANet achieved higher performance on common indicators. For example, on the CT dataset, the PSNR and SSIM indicators were improved by 4.1dB and 0.06 respectively, and the rMSE indicator was reduced by 5.46 compared with the LDCT; on the PET dataset, the PSNR and SSIM was improved by 3.37dB and 0.02, and the rMSE was reduced by 9.04 compared with theLDPET. This paper proposes a transformer-based denoising algorithm, which utilizes hybrid attention to extract high-level features of low dose images and fuses noise features to optimize the denoising performance of the network, achieving good performance improvements on low-dose CT and PETdatasets.

Three-dimentional (3D) vessel wall magnetic resonance imaging (MRI) sequences have emerged as new imaging tools for evaluating carotid atherosclerosis. However, their reproducibility across different vendors has not yet been investigated, which not only restricts their use in multicenter studies but also hinders their broader application in clinical practice. In this study, we aim to assess the qualitative and quantitative reproducibility on the same subjects using matched 3D carotid vessel wall MRI sequences on both Siemens and Philips scanners, specifically, 3D motion-sensitized driven equilibrium prepared rapid gradient echo (MERGE) and simultaneous non-contrast angiography and plaque (SNAP) imaging which are two representative 3D vessel wall MRI sequences with superior delineation of vessel wall morphology and carotid plaque. As a cross-sectional study, six volunteers (1 female and 5 males, age 22-67 years) were scanned at 3T MRI machines of both vendors. Image quality was evaluated by two experienced reviewers using a 4-point scale, and quantitative measurements, including mean/maximum wall thickness and normalized wall/lumen index, were calculated from segmentation masks generated by the 3D localization, analysis, and thickness and tissue evaluation (LATTE) framework and a novel 3D thickness measurement using Laplacian method. There was no significant difference in image quality scores between Siemens and Philips platforms, except in the external carotid artery region. High consistency [intra-class correlation coefficient (ICC) >0.75] was obtained between the two platforms for quantitative metrics. Images on one carotid patient on Siemens show good visualization of vessel wall and plaque morphology and detection of intraplaque hemorrhage. 3D MERGE and SNAP images have sufficient image quality and consistent quantitative measurements on Siemens and Philips scanners, despite lower image quality in Siemens platforms, probably due to suboptimal coil configuration or image processing. This suggests the feasibility of evaluating carotid atherosclerosis using matched 3D carotid vessel wall MRI protocols across different MRI vendors.

One-stage pulmonary nodule detection using 3-D DCNN with feature fusion and attention mechanism in CT image

In low-dose computed tomography (LDCT) denoising tasks, it is often difficult to balance edge/detail preservation and noise/artifact reduction. To solve this problem, we propose a dual convolutional neural network (CNN) based on edge feature extraction (Ed-DuCNN) for LDCT. Ed-DuCNN consists of two branches. One branch is the edge feature extraction subnet (Edge_Net) that can fully extract the edge details in the image. The other branch is the feature fusion subnet (Fusion_Net) that introduces an attention mechanism to fuse edge features and noisy image features. Specifically, first, shallow edge-specific detail features are extracted by trainable Sobel convolutional blocks and then are integrated into Edge_Net together with the LDCT images to obtain deep edge detail features. Finally, the input image, shallow edge detail, and deep edge detail features are fused in Fusion_Net to generate the final denoised image. The experimental results show that the proposed Ed-DuCNN can achieve competitive performance in terms of quantitative metrics and visual perceptual quality compared with that of state-of-the-art methods.

Since low-dose computed tomography (LDCT) images typically have higher noise that may affect accuracy of disease diagnosis, the objective of this study is to develop and evaluate a new artifact-assisted feature fusion attention (AAFFA) network to extract and reduce image artifact and noise in LDCT images. In AAFFA network, a feature fusion attention block is constructed for local multi-scale artifact feature extraction and progressive fusion from coarse to fine. A multi-level fusion architecture based on skip connection and attention modules is also introduced for artifact feature extraction. Specifically, long-range skip connections are used to enhance and fuse artifact features with different depth levels. Then, the fused shallower features enter channel attention for better extraction of artifact features, and the fused deeper features are sent into pixel attention for focusing on the artifact pixel information. Besides, an artifact channel is designed to provide rich artifact features and guide the extraction of noise and artifact features. The AAPM LDCT Challenge dataset is used to train and test the network. The performance is evaluated by using both visual observation and quantitative metrics including peak signal-noise-ratio (PSNR), structural similarity index (SSIM) and visual information fidelity (VIF). Using AAFFA network improves the averaged PSNR/SSIM/VIF values of AAPM LDCT images from 43.4961, 0.9595, 0.3926 to 48.2513, 0.9859, 0.4589, respectively. The proposed AAFFA network is able to effectively reduce noise and artifacts while preserving object edges. Assessment of visual quality and quantitative index demonstrates the significant improvement compared with other image denoising methods.

X-ray computed tomography (CT) is vital for medical diagnostics, but frequent radiation exposure raises concerns, driving the adoption of low-dose CT (LDCT) to mitigate risks. However, LDCT often introduces noise, compromising diagnostic accuracy. This paper proposes a pure vision transformer (PViT) for LDCT denoising, enhanced with a gradient–Laplacian attention module (GLAM) to improve edge preservation and fine structural detail reconstruction. The model’s robustness was validated across five diverse datasets (piglet, head, abdomen, chest, thoracic), demonstrating consistent performance in preserving anatomical structures. Extensive ablation studies on attention configurations and loss functions further substantiated the contributions of each module. Quantitative evaluation using PSNR and SSIM, alongside radiologist assessment, confirmed significant noise suppression and sharper anatomical boundaries, particularly in regions with fine details such as organ interfaces and bone structures. Additionally, in benchmark comparisons against state-of-the-art LDCT models (RED-CNN, TED-Net, DSC-GAN, DRL-EMP) and traditional methods (BM3D), the model exhibited lower parameter and stable training performance. These findings highlight the model’s robustness, efficiency, and clinical applicability, making it a promising solution for improving LDCT image quality while maintaining computational efficiency.

A coarse-to-fine multi-scale feature hybrid low-dose CT denoising network

GDAFormer: Gradient-guided Dual Attention Transformer for Low-Dose CT image denoising

With the increasing concern regarding the radiation exposure of patients undergoing computed tomography (CT) scans, researchers have been using deep learning techniques to improve the quality of denoised low-dose CT (LDCT) images. In this paper, a cascaded dilated residual network (ResNet) with integrated attention modules, specifically spatial- and channel- attention modules, is proposed. This experiment demonstrated how these attention modules improved the denoised CT image by testing a simple ResNet with and without the modules. Further, an investigation regarding the effectiveness of per-pixel loss, perceptual loss via VGG16-Net, and structural dissimilarity loss functions is also covered through an ablation experiment. By knowing how these loss functions affect the output denoised images, a combination of the these loss function is then proposed which aims to prevent edge over-smoothing, enhance textural details and finally, preserve structural details on the denoised images. Finally, a bench testing was also done by comparing the visual and quantitative results of the proposed model with the state-of-the-art models such as block matching 3D (BM3D), patch-GAN and dilated convolution with edge detection layer (DRL-E-MP) for accuracy.

In some noisy low dose CT lung cancer screening images, we noticed that the CT density values of air were increased and the visibility of emphysema was distinctly decreased. By examining histograms of these images, we found that the CT density values were truncated at -1024 HU. The purpose of this study was to investigate the effect of pixel value truncation on the visibility of emphysema using mathematical models. Assuming CT noise follows a normal distribution, we derived the relationship between the mean CT density value and the standard deviation (SD) when the pixel values below -1024 HU are truncated and replaced by -1024 HU. To validate our mathematical model, 20 untruncated phantom CT images were truncated by simulation, and the mean CT density values and SD of air in the images were measured and compared with the theoretical values. In addition, the mean CT density values and SD of air were measured in 100 cases of real clinical images obtained by GE, Siemens, and Philips scanners, respectively, and the agreement with the theoretical values was examined. Next, the contrast-to-noise ratio (CNR) between air (-1000 HU) and lung parenchyma (-850 HU) was derived from the mathematical model in the presence and absence of truncation as a measure of the visibility of emphysema. In addition, the radiation dose ratios required to obtain the same CNR in the case with and without truncation were also calculated. The mathematical model revealed that when the pixel values are truncated, the mean CT density values are proportional to the noise magnitude when the magnitude exceeds a certain level. The mean CT density values and SD measured in the images with pixel values truncated by simulation and in the real clinical images acquired by GE and Philips scanners agreed well with the theoretical values from our mathematical model. In the Siemens images, the measured and theoretical values agreed well when a portion of the truncated values were replaced by random values instead of simply replacing by -1024 HU. The CNR of air and lung parenchyma was lowered by truncating CT density values compared to that of no truncation. Furthermore, it was found that higher radiation dose was required to obtain the same CNR with truncation as without. As an example, when the noise SD was 60 HU, the radiation dose required for the GE and Philips truncation method was about 1.2 times higher than that without truncation, and that for the Siemens truncation method was about 1.4 times higher. It was demonstrated mathematically that pixel value truncation causes a brightening of the mean CT density value and decreases the CNR of emphysema. Our results indicate that it is advisable to turn off truncation at -1024 HU, especially when scanning at low and ultra-low radiation doses in the thorax.

Low-dose computed tomography (LDCT) denoising is an indispensable procedure in the medical imaging field, which not only improves image quality, but can mitigate the potential hazard to patients caused by routine doses. Despite the improvement in performance of the cycle-consistent generative adversarial network (CycleGAN) due to the well-paired CT images shortage, there is still a need to further reduce image noise while retaining detailed features. Inspired by the residual encoder–decoder convolutional neural network (RED-CNN) and U-Net, we propose a novel unsupervised model using CycleGAN for LDCT imaging, which injects a two-sided network into selective kernel networks (SK-NET) to adaptively select features, and uses the patchGAN discriminator to generate CT images with more detail maintenance, aided by added perceptual loss. Based on patch-based training, the experimental results demonstrated that the proposed SKFCycleGAN outperforms competing methods in both a clinical dataset and the Mayo dataset. The main advantages of our method lie in noise suppression and edge preservation.

Low-dose computed tomography (CT) denoising algorithms aim to enable reduced patient dose in routine CT acquisitions while maintaining high image quality. Recently, deep learning (DL)-based methods were introduced, outperforming conventional denoising algorithms on this task due to their high model capacity. However, for the transition of DL-based denoising to clinical practice, these data-driven approaches must generalize robustly beyond the seen training data. We, therefore, propose a hybrid denoising approach consisting of a set of trainable joint bilateral filters (JBFs) combined with a convolutional DL-based denoising network to predict the guidance image. Our proposed denoising pipeline combines the high model capacity enabled by DL-based feature extraction with the reliability of the conventional JBF. The pipeline’s ability to generalize is demonstrated by training on abdomen CT scans without metal implants and testing on abdomen scans with metal implants as well as on head CT data. When embedding RED-CNN/QAE, two well-established DL-based denoisers in our pipeline, the denoising performance is improved by 10%/82% (RMSE) and 3%/81% (PSNR) in regions containing metal and by 6%/78% (RMSE) and 2%/4% (PSNR) on head CT data, compared to the respective vanilla model. Concluding, the proposed trainable JBFs limit the error bound of deep neural networks to facilitate the applicability of DL-based denoisers in low-dose CT pipelines.We made our trainable bilateral filter layer package (PyTorch, GPU accelerated) publicly available [1, 2].

HPIDN: A Hierarchical prior-guided iterative denoising network with global–local fusion for enhancing low-dose CT images

<p>Radiology is a critical tool for physicians when treating many health complications. X-ray computed tomography (X-ray CT) is particularly useful in highlighting lesions and damaged tissue in the body. The adoption of X-ray tomography is seeing significant growth as the technology improves and radiologists are finding it increasingly useful. A significant drawback of X-ray CT is the associated radiation exposure which is harmful to the body. If the radiation dose is decreased in the acquisition protocol, the images become degraded and no longer useful for radiologists. Current methods for reducing radiation risk include guidelines on acceptable exposure, and very recently reconstruction techniques which aim to denoise images acquired at low radiation dose. However, image post-processing techniques are preferred because they are scanner independent, and don’t require the denoising algorithm to have access to the raw scanner data which is often not made available by manufacturers. With advances in deep-learning via convolutional neural networks, several methods have been proposed to denoise low dose CT images in order to predict the normal dose image. In this work, existing methods are improved on with the addition of our contributions. We propose a training regime using a cascade of neural networks the first of which uses a perceptual loss, and the second which performs a residue prediction using mean-squared-error. Secondly, we propose a new loss function which incorporates perceptual loss, structural dissimilarity and mean-squared-error. We show that both proposed methods result in significant improvement when compared to the related works. As a secondary contribution, this work recommends several design considerations for building deep learning networks for image denoising. We show results from extensive empirical study to support our recommendations.</p>