| CT(Computed tomography)is a widely used and important clinical diagnosis tool.However,the radiation of CT greatly increases the carcinogenic risk of the human body.The radiation dose must lower in order to reduce the harm of X-rays to the human body.The low-dose-radiation X-ray will degrade the clinical diagnosis performance because it will induce artifacts and noise in CT image reconstruction.Even with normal radiation dose imaging,CT images still have some unavoidable artifacts: such as motion artifacts,metal artifacts,etc.By traditional CT technology alone,it is impossible to obtain perfect CT images.However,it is possible to effectively remove noise and artifacts from CT images by using recently developed image-processing algorithms,thereby enhancing the details and quality of the image.Therefore,this paper proposes a software algorithm to remove artifacts from CT images.On the other hand,in recent years,with the enhancing computing power of hardware devices,the success of in-depth learning in computer vision has made it widely studied and applied in various fields.The achievements of in-depth learning in image processing have attracted the attention of reasercher from medical imaging.In-depth learning technology is increasingly applied to medical image processing.In this paper,we focus on using the algorithm of GANs(Generative Adversarial Networks)in-depth learning to eliminate the impairment of artifacts on CT images quality.The main research work and results are as follows:(1)This paper proposes to use the neural network algorithm CT-WGAN based on the GANs to process the artifacts in the medical image and restore the details of the CT image to improve the image quality.Recently,in-depth learning primarily use traditional in-depth lerning convolutional neural networks for medical image processing.In contras,in-depth learning GAN proposed in this paper is seldom studied,thus the study in this paper on GAN is meaningful for potential application on CT imaging.(2)This paper discusses the technology based on the GANs,including the latest proposed technologies,including the use of MobileNetV2 network framework,linear inverted residual network structure,WGAN loss function and WGAN training enhancement,thus gains more better processing result than that by using traditional neural network algorithms.(3)OpenVINO toolkit is used to optimize the code for CT-WGAN and thus accelerate the speed of medical CT processing.It can decrease the requirement and the cost of equiment for medical CT image processing,therefore common computer available in market can be use for this purpose.In conclusion,the proposed CT-WGAN realizes the CT de-artifact function and has practical application and reference value. |
