Yuewen Sun

Bio: Yuewen Sun is an academic researcher from Tsinghua University. The author has contributed to research in topics: Nuclear engineering & Graphite. The author has an hindex of 3, co-authored 3 publications receiving 34 citations.

Digital radiography image denoising using a generative adversarial network.

Abstract: Statistical noise may degrade the x-ray image quality of digital radiography (DR) system. This corruption can be alleviated by extending exposure time of detectors and increasing the intensity of radiation. However, in some instances, such as the security check and medical imaging examination, the system demands rapid and low-dose detection. In this study, we propose and test a generative adversarial network (GAN) based x-ray image denoising method. Images used in this study were acquired from a digital radiography (DR) imaging system. Promising results have been obtained in our experiments with x-ray images for the security check application. The Experiment results demonstrated that the proposed new image denoising method was able to effectively remove the statistical noise from x-ray images, while kept sharp edge and clear structure. Thus, comparing with the traditional convolutional neural network (CNN) based method, the proposed new method generates more plausible-looking images, which contains more details.

Single-image super-resolution reconstruction method based on deep residual network

Abstract: The invention discloses a single-image super-resolution reconstruction method based on a deep residual network. The method of the invention mainly comprises a first step of performing block extraction and pixel averaging processing on an image in a sample image database to obtain a corresponding high resolution and low resolution training image sets; a second step of constructing a deep convolutional neutral network with a residual structure for iterative training, and then inputting the training set obtained in the first step to the neural network constructed in the second step for iterative training; and a third step of according to a data model obtained by training, realizing the continuous up-scaling of the input low resolution image through the combination of iterative operation and an interpolation algorithm. By introducing a deep residual network and introducing an upsampling layer at the end of the network, the method of the invention accelerates the processing speed of the image up-scaling, enhances the display effect of the image details, obtains a better image super-resolution reconstruction effect, and has a wide range of applications in the image high definition display, image compression, security checks and other fields.

Enhancement of digital radiography image quality using a convolutional neural network

Abstract: Digital radiography system is widely used for noninvasive security check and medical imaging examination. However, the system has a limitation of lower image quality in spatial resolution and signal to noise ratio. In this study, we explored whether the image quality acquired by the digital radiography system can be improved with a modified convolutional neural network to generate high-resolution images with reduced noise from the original low-quality images. The experiment evaluated on a test dataset, which contains 5 X-ray images, showed that the proposed method outperformed the traditional methods (i.e., bicubic interpolation and 3D block-matching approach) as measured by peak signal to noise ratio (PSNR) about 1.3 dB while kept highly efficient processing time within one second. Experimental results demonstrated that a residual to residual (RTR) convolutional neural network remarkably improved the image quality of object structural details by increasing the image resolution and reducing image noise. Thus, this study indicated that applying this RTR convolutional neural network system was useful to improve image quality acquired by the digital radiography system.

Defect Detection Method for Large Size Graphite and Carbon Components of High Temperature Gas Cooled Reactor Based on Computed Tomography

Abstract: High temperature gas cooled reactor (HTGR) is a typical type of the fourth-generation nuclear power system. The main supporting structure, consisting of graphite and carbon components, play a vital role in the construction of the HTGR. The quality of the components is essential for the safety operation of HTGR since they are irreplaceable throughout the reactor lifetime. The manufacture of the components is complex, including multiple process, during which defects such as holes and crack often arise inevitably. These defects may bring serious risk to the structural safety and steady operation of the reactor. Therefore, it is of great significance to inspect and evaluate the quality of the components. Considering the large size of the components as well as the long production cycle, traditional non-destructive testing method such as x-ray and eddy current testing are not applicable. Visual inspection and spot check are generally applied to check the surface condition, which are unable to provide the internal situation of the components. This paper proposes a helical CT based defects detection method for large size graphite and carbon components in HTGR. Graphite and carbon samples with artificial and original defect were produced, and various experiments were conducted on a multi-slice helical CT system to check the performance as well as optimize the operation parameter. The results indicates that defect larger than 2 mm in graphite components and 1 mm in carbon components can be detected and clearly visualized, which proves the feasibility of the proposed method.

Monte Carlo Simulation of In-Situ Multi-Scale Damage Detection of AGR Graphite

Abstract: The Advanced Gas-Cooled Reactor (AGR) is the most common design of nuclear reactor in the UK. These reactors were commissioned in the 1980s and have now reached the end of their design life. As a type of thermal reactor, it uses graphite as a moderator to control the reaction. The cracking of graphite blocks in the core is an important factor limiting the lifetime of AGR stations. The damage of the AGR graphite will affect the power of the reactor and also become a hidden danger threatening the safety of the reactor. As a thermal reactor, it uses graphite as a moderator. Since graphite bricks cannot be replaced or repaired, in order to ensure the continued operation of the reactor, it is necessary to obtain timely and accurate damage information of graphite bricks. In order to carry out the research on in-situ multi-scale damage detection of AGR graphite, a linear CT imaging detection scheme was designed in combination with the core geometries of the AGR stations. Based on the GEANT4 Monte Carlo simulation software, an experiment is carried out to simulate a series of processes from signal acquisition to image reconstruction. Through the simulation experiment, the detection ability of the linear CT detection system under the interference of strong background radiation field signal in the AGR reactor and the geometric error of the detection system is verified, so as to evaluate the feasibility of the linear CT imaging detection scheme and guide the design of the detection system.

Identification of COVID-19 samples from chest X-Ray images using deep learning: A comparison of transfer learning approaches.

Abstract: Background The novel coronavirus disease 2019 (COVID-19) constitutes a public health emergency globally. The number of infected people and deaths are proliferating every day, which is putting tremendous pressure on our social and healthcare system. Rapid detection of COVID-19 cases is a significant step to fight against this virus as well as release pressure off the healthcare system. Objective One of the critical factors behind the rapid spread of COVID-19 pandemic is a lengthy clinical testing time. The imaging tool, such as Chest X-ray (CXR), can speed up the identification process. Therefore, our objective is to develop an automated CAD system for the detection of COVID-19 samples from healthy and pneumonia cases using CXR images. Methods Due to the scarcity of the COVID-19 benchmark dataset, we have employed deep transfer learning techniques, where we examined 15 different pre-trained CNN models to find the most suitable one for this task. Results A total of 860 images (260 COVID-19 cases, 300 healthy and 300 pneumonia cases) have been employed to investigate the performance of the proposed algorithm, where 70% images of each class are accepted for training, 15% is used for validation, and rest is for testing. It is observed that the VGG19 obtains the highest classification accuracy of 89.3% with an average precision, recall, and F1 score of 0.90, 0.89, 0.90, respectively. Conclusion This study demonstrates the effectiveness of deep transfer learning techniques for the identification of COVID-19 cases using CXR images.

Vision Transformer based COVID-19 Detection using Chest X-rays

Abstract: COVID-19 is a global pandemic, and detecting them is a momentous task for medical professionals today due to its rapid mutations. Current methods of examining chest X-rays and CT scan requires profound knowledge and are time consuming, which suggests that it shrinks the precious time of medical practitioners when people’s lives are at stake. This study tries to assist this process by achieving state-of-the-art performance in classifying chest X-rays by fine-tuning Vision Transformer(ViT). The proposed approach uses pretrained models, fine-tuned for detecting the presence of COVID-19 disease on chest X-rays. This approach achieves an accuracy score of 97.61%, precision score of 95.34%, recall score of 93.84% and, fl-score of 94.58%. This result signifies the performance of transformer-based models on chest X-ray.

Comparison of medical image classification accuracy among three machine learning methods

Abstract: Background Low-quality medical images may influence the accuracy of the machine learning process. Objective This study was undertaken to compare accuracy of medical image classification among machine learning methods, as classification is a basic aspect of clinical image inspection. Methods Three types of machine learning methods were used, which include Support Vector Machine (SVM), Artificial Neural Network (ANN), and Convolution Neural Network (CNN). To investigate changes in accuracy related to image quality, we constructed a single dataset using two different file formats of DICOM (Digital Imaging and Communications in Medicine) and JPEG (Joint Photographic Experts Group). Results The JPEG format contains less color information and data capacity than the DICOM format. CNN classification was accurate for both datasets, whereas SVM and ANN accuracy decreased with the loss of data from DICOM to JPEG formats. Conclusions CNN is more accurate than conventional machine learning methods that utilize the manual feature extraction.