Article
Radiology, Nuclear Medicine & Medical Imaging
Xiaohuan Yu, Ailong Cai, Lei Li, Zhiyong Jiao, Bin Yan
Summary: This study developed a method that jointly utilizes global, local, and nonlocal priors to develop the reconstruction model for low-dose MECT. The global low-rankness and nonlocal prior are cascaded using subspace decomposition and block-matching, while the L0 sparsity is applied to express the local prior. Experimental results demonstrate that the proposed method has advantages in computational efficiency, noise suppression, and structure preservation.
QUANTITATIVE IMAGING IN MEDICINE AND SURGERY
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Mayank Patwari, Ralf Gutjahr, Rainer Raupach, Andreas Maier
Summary: In this study, a novel CT denoising framework is introduced, which employs bilateral filtering in both the projection and volume domains to remove noise. Two deep CNNs are used for parameter tuning and training via a reward network. The framework demonstrates excellent denoising performance and outperforms other deep CNN models. It does not introduce blurring or deep learning artifacts.
Article
Computer Science, Artificial Intelligence
Jiadong Zhang, Zhiming Cui, Caiwen Jiang, Shanshan Guo, Fei Gao, Dinggang Shen
Summary: This article proposes a learning-based method to reconstruct high-dose positron emission tomography (PET) images from low-dose PET images and corresponding total-body computed tomography (CT) images. The proposed hierarchical framework can consistently improve the performance of all body parts and outperforms the state-of-the-art methods in single-photon emission computed tomography (SPET) image reconstruction, with a peak signal-to-noise ratio (PSNR) of 30.6 dB for total-body PET images.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Dufan Wu, Kyungsang Kim, Quanzheng Li
Summary: The proposed deep learning-based low-dose CT image reconstruction algorithm, utilizing Noise2Noise network, does not require clean images for training, achieving better performance compared to existing methods. The method is robust against different noise levels, hyperparameters, and network structures, and can achieve competitive results without any pre-training of the network weights. The iterative reconstruction algorithm also shows empirical convergence with and without network pre-training.
Article
Engineering, Electrical & Electronic
Wei Yu, Wei Peng, Hai Yin, Chengxiang Wang, Kaihu Yu
Summary: The proposed joint regularization method combines structural group sparsity and gradient prior sparsity to solve the optimization-based CT reconstruction problem. It shows better performance than other iterative reconstruction algorithms in both simulated experiments and real data studies, effectively reconstructing important structure features and suppressing noise and artifacts.
Article
Biochemical Research Methods
Han Liu, Peixi Liao, Hu Chen, Yi Zhang
Summary: Computed tomography (CT) is a powerful tool for medical diagnosis, but obtaining suitable low dose CT images while minimizing X-ray radiation risk poses challenges. This paper proposes ERA-WGAT, a method that incorporates an edge enhancement module and a window-based graph attention convolutional network to extract non-local information and suppress noise, achieving superior results.
BIOMEDICAL OPTICS EXPRESS
(2022)
Article
Engineering, Electrical & Electronic
Haijun Yu, Shaoyu Wang, Weiwen Wu, Changcheng Gong, Linbo Wang, Zhenzhen Pi, Fenglin Liu
Summary: A weighted adaptive non-local dictionary (WAND) method has been developed for low-dose computed tomography (LDCT) image reconstruction, achieving better image quality with small details preservation and noise suppression. By iteratively characterizing the noise property in each local patch and adaptively describing the different distributions of sparse coefficients of each patch, WAND shows promising performance in LDCT reconstruction.
Article
Engineering, Electrical & Electronic
Shaoyu Wang, Haijun Yu, Yarui Xi, Changcheng Gong, Weiwen Wu, Fenglin Liu
Summary: In this study, a spectral-image decomposition with energy-fusion sensing (SIDES) reconstruction method is proposed to obtain better quality spectral images and material decomposition results by establishing a unified tensor decomposition model. The method fully explores spatial sparsity, global correlation across the spectrum, and nonlocal self-similarity properties, and outperforms the state-of-the-art in numerical simulations and real experiments.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Andrea Agostini, Alessandra Borgheresi, Marina Carotti, Letizia Ottaviani, Myriam Badaloni, Chiara Floridi, Andrea Giovagnoni
Summary: The study showed that low-dose CT scans with third-generation iterative reconstruction technology provided acceptable image quality for evaluating COVID-19 patients, significantly reducing radiation dose and motion artifacts.
Article
Computer Science, Interdisciplinary Applications
Wenjun Xia, Zexin Lu, Yongqiang Huang, Yan Liu, Hu Chen, Jiliu Zhou, Yi Zhang
Summary: This paper introduces a parameter-dependent framework that trains a reconstruction network with data from multiple alternative geometries and dose levels simultaneously, reducing extra training costs for multiple geometries and dose levels.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Yuanwei He, Li Zeng, Wei Chen, Changcheng Gong, Zhaoqiang Shen
Summary: In this study, a BRTV-regularized POCS model is proposed for LDCT reconstruction, which can simultaneously maintain edges and reduce noise, resulting in superior image quality.
JOURNAL OF DIGITAL IMAGING
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Salim A. Si-Mohamed, Joel Greffier, Jade Miailhes, Sara Boccalini, Pierre-Antoine Rodesch, Aurelie Vuillod, Niels van der Werf, Djamel Dabli, Damien Racine, David Rotzinger, Fabio Becce, Yoad Yagil, Philippe Coulon, Alain Vlassenbroek, Loic Boussel, Jean-Paul Beregi, Philippe Douek
Summary: The study evaluated the image quality of spectral photon-counting CT (SPCCT) compared to dual-layer CT (DLCT) with different reconstruction algorithms. Results showed that SPCCT had lower noise magnitude and higher detectability for nodules compared to DLCT. SPCCT provided higher image quality and better conspicuity for both ground-glass nodules and solid nodules at different iDose(4) levels.
EUROPEAN RADIOLOGY
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Lingming Zeng, Xu Xu, Wen Zeng, Wanlin Peng, Jinge Zhang, Sixian Hu, Keling Liu, Chunchao Xia, Zhenlin Li
Summary: The study found that using DELTA for half-dose contrast-enhanced liver CT scans can significantly reduce dosage by approximately 49% compared to using HIR for standard-dose CT, while maintaining image quality.
EUROPEAN JOURNAL OF RADIOLOGY
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Hye Joo Park, Seo-Youn Choi, Ji Eun Lee, Sanghyeok Lim, Min Hee Lee, Boem Ha Yi, Jang Gyu Cha, Ji Hye Min, Bora Lee, Yunsub Jung
Summary: This study compared the image quality and radiation dose of a deep learning image reconstruction algorithm (DLIR) with iterative reconstruction (IR) and filtered back projection (FBP) at different tube voltages and tube currents. The results showed that DLIR significantly reduced noise and artifacts and improved overall image quality compared to FBP and hybrid IR. Despite the reduced image sharpness, low-dose CT with DLIR seemed to have a greater potential for dose optimization.
EUROPEAN RADIOLOGY
(2022)
Article
Computer Science, Artificial Intelligence
Lianying Chao, Peng Zhang, Yanli Wang, Zhiwei Wang, Wenting Xu, Qiang Li
Summary: A dual-domain attention-guided network framework (Dual-AGNet) is developed to improve the quality of low-dose cone-beam computed tomography (CBCT) images. This method processes images in both projection and reconstruction domains, and achieves superior performance while recovering lost structures.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Engineering, Electrical & Electronic
Tao Zhang, Haijun Yu, Yarui Xi, Shaoyu Wang, Fenglin Liu
Summary: The spectral computed tomography (CT) system with a photon-counting detector (PCD) can analyze the material composition of the inspected object through material decomposition. However, the raw projection of spectral CT is often affected by noise and artifacts, leading to poor quality material decomposition images. To overcome this limitation, a sparsity residual prior and dictionary learning (SRPDL) algorithm was proposed, which improves noise reduction and edge protection in material decomposition.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Engineering, Electrical & Electronic
Weiwen Wu, Xiaodong Guo, Yang Chen, Shaoyu Wang, Jun Chen
Summary: Tomographic image reconstruction with deep learning is a new field of applied artificial intelligence. The task of reducing radiation dose with sparse views' reconstruction is significant in cardiac imaging. This study proposes a DEAR network to address the challenge of obtaining good images from high sparse-view level, with modules for deep embedding, deep attention, and deep refinement. The results on clinical datasets demonstrate the efficiency of DEAR in edge preservation and feature recovery.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Shaoyu Wang, Weiwen Wu, Ailong Cai, Yongshun Xu, Varut Vardhanabhuti, Fenglin Liu, Hengyong Yu
Summary: In this paper, we propose an image-spectral decomposition extended-learning assisted by sparsity (IDEAS) method for multi-energy CT image reconstruction. This method effectively utilizes the intrinsic priors of multi-energy CT images through nonlocal low-rank Tucker decomposition and multi-task tensor dictionary learning. Numerical simulation, physical phantom, and preclinical mouse experiments demonstrate the outstanding performance of the proposed IDEAS algorithm.
QUANTITATIVE IMAGING IN MEDICINE AND SURGERY
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Shaoyu Wang, Ailong Cai, Weiwen Wu, Tao Zhang, Fenglin Liu, Hengyong Yu
Summary: This study develops an image-domain material decomposition method (IMD-MTFC) for spectral CT, which utilizes material-image tensor factorization and clustering to achieve high-precision material-specific imaging.
IEEE TRANSACTIONS ON RADIATION AND PLASMA MEDICAL SCIENCES
(2023)
Article
Engineering, Multidisciplinary
Haijun Yu, Song Ni, Jie Chen, Wenjie Ge, Lingli Zhang, Fenglin Liu
Summary: In this study, an analytical algorithm is proposed to address truncated projections, redundant data, and geometry errors in the reconstruction process of multiple X-ray source translation micro-CT (mSTCT). Truncated projections are handled by introducing virtual projections, redundant data are dealt with using a new weighting strategy, and geometry errors are mitigated by modifying the projection geometry. Numerical and physical experiments confirm the effectiveness of the proposed algorithm for mSTCT reconstruction.
APPLIED MATHEMATICAL MODELLING
(2023)
Article
Computer Science, Interdisciplinary Applications
Chuang Niu, Mengzhou Li, Fenglei Fan, Weiwen Wu, Xiaodong Guo, Qing Lyu, Ge Wang
Summary: Image denoising is crucial for downstream tasks in various fields. The proposed self-supervised denoising approach, Noise2Sim, is capable of suppressing both independent and correlated noises, which current methods fail to handle. The experimental results show that Noise2Sim performs as effectively as, or even better than, supervised learning methods in recovering intrinsic features from noisy CT images. Noise2Sim is a versatile self-supervised denoising approach with great potential for diverse applications.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Article
Optics
Song Ni, Haijun Yu, Jie Chen, Chuanjiang Liu, Fenglin Liu
Summary: This paper introduces a hybrid source translation scanning mode, called hySTCT, for interior tomography, which can efficiently image large objects. By finely sampling the projections inside and coarsely sampling the projections outside the region of interest (ROI), the truncation artifacts and value bias in interior tomography can be reduced. Two reconstruction methods, interpolation V-FBP (iV-FBP) and two-step V-FBP (tV-FBP), are developed based on the linearity property of the inverse Radon transform for hySTCT reconstruction. The experiments demonstrate that the proposed strategy can effectively suppress truncation artifacts and improve reconstruction accuracy within the ROI.
Article
Environmental Sciences
Ying Chu, Fan Chen, Hong Fu, Hengyong Yu
Summary: Air quality monitoring is important for human health, but traditional methods have limitations in complex terrains and high costs. This paper proposes a novel idea of using visual assessment of haze in photos to monitor air pollution in urban areas. The correlation between air quality indexes and haziness scores of monitoring images is calculated, showing that objective indicators can reflect air pollution levels. A new method called fastDBCP is developed to apply this observation practically, which demonstrates advantages in terms of correlation degree, computational speed, and classification accuracy.
Article
Health Care Sciences & Services
Tsung-Hsien Yu, Frank Leigh Lu, Chung-Jen Wei, Wei-Wen Wu
Summary: This study used data from Taiwan's National Health Insurance claims, Death Registry, and Cancer Registry to investigate the impact of reimbursement policy expansion on hospice care use among populations with different demographics characteristics and health status. The results showed an increasing trend in hospice care use after the scope of benefits expansion, but no increase in the initiation time of first hospice care use. The effects of expansion varied among patients based on their demographic characteristics.
BMC PALLIATIVE CARE
(2023)
Article
Chemistry, Multidisciplinary
Fan Chen, Hong Fu, Hengyong Yu, Ying Chu
Summary: A no-reference image quality assessment method is proposed based on a multitask image restoration network. The method utilizes the internal generative mechanism of the human visual system to evaluate image quality by integrating both local and global information, restoring the original image information using contextual information, and comparing it with the distorted image information. Experimental results demonstrate excellent performance on both synthetically and authentically distorted databases.
APPLIED SCIENCES-BASEL
(2023)
Article
Chemistry, Analytical
Fan Chen, Hong Fu, Hengyong Yu, Ying Chu
Summary: In this paper, a dual-pathway convolutional neural network is proposed for blind image quality assessment. The proposed method extracts content features and global shape features of the image by mimicking the ventral pathway and dorsal pathway of the human visual system, respectively. The features from the two pathways are then fused and mapped to an image quality score, achieving state-of-the-art performance according to experiments on multiple databases.
Article
Engineering, Electrical & Electronic
Weiwen Wu, Jianjia Zhang, Shaoyu Wang, Jun Chen
Summary: A new stationary CT scanner called FANS, based on a flat-panel addressable nanowire-cold-cathode architecture, is proposed. The FANS scanner is a small and cheap system as it utilizes a low-cost compact flat-panel X-ray source. It is also a stationary CT system with fast-switching capabilities. Moreover, the FANS system provides high-quality image output by incorporating superior regularization priors.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Engineering, Electrical & Electronic
Xiaodong Guo, Yonghui Li, Dingyue Chang, Peng He, Peng Feng, Hengyong Yu, Weiwen Wu
Summary: The paper proposes an iterative deep reconstruction network, Spectral2Spectral, which combines unsupervised methods and data priors into a unified framework to obtain high-quality images from noisy data in an end-to-end fashion. The structural similarity prior in the image-spectral domain is refined as a regularization term to further constrain network training. The weights of the neural network are automatically updated to capture image features and structures within the iterative process.
IEEE TRANSACTIONS ON COMPUTATIONAL IMAGING
(2023)
Article
Engineering, Electrical & Electronic
Jie Chen, Haijun Yu, Song Ni, Chuanjiang Liu, Wenjie Ge, Yixing Huang, Fenglin Liu
Summary: Micro-computed tomography (micro-CT) is an important tool for providing high-resolution 3D images in scientific research. This article introduces a multiple source translation computed tomography (mSTCT) imaging geometry and virtual projection-based filtered back-projection (V-FBP) reconstruction algorithm, and proposes a full-scan multiple-STCT (F-mSTCT) scanning configuration and a weighted FBP (W-FBP) algorithm, which can achieve high-resolution reconstruction at a low source sampling rate.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)