Research On Low Dose Liver And Lung Denoising Algorithm

Computer Tomograpghy,as a commonly used clinical imaging technique,has the advantages of fast scanning and clear images,but the impact of ionizing radiation carried in standard dose CT scans on the human body cannot be ignored,which also leads to the current low Dose CT scans are becoming more and more popular.Lowdose CT ionizing radiation will be less than normal dose CT,but the noise contained therein will greatly affect the doctor’s judgment.Therefore,the content of this subject is to use deep learning technology to reduce the noise contained in low-dose CT.The main research work of this paper is as follows:First of all,this paper proposes a noise extraction method based on multiple attention mechanisms.This method can better find the prior information of noise distribution from low-dose CT images,which is very important for subsequent image generation.Link.Experiments show that if the normal convolutional neural network is used to extract the noise distribution,the effect of extracting the noise distribution is not ideal,so three modules of spatial attention,channel attention and scale attention are added to the network.Experiments show that this method can better extract noise distribution information.Secondly,for the calculation method of perceptual loss,this paper uses the retrained Autoencoder network to extract the features of the normal dose CT images in the data set used in this subject.Experiments show that this network can guide the network to generate better final images.Finally,based on the two network modules mentioned above and the improved Generative Adversarial Networks,this paper proposes a final network structure for low-dose CT noise reduction.The noise extraction module is added to the generator of Generative Adversarial Networks,and the generator is the perceptual loss of the encoder part based on Autoencoder is added to the loss.The experimental results show that the experimental results have good performance.