Processing And Comprehensive Analysis Of Lung Medical CT Image Based On Deep Learning

As of January 2022,there were more than 285 million confirmed cases of COVID-19 worldwide,affecting almost every region of the world.The severe epidemic situation is testing the level of medical technology and the ability to respond to medical emergencies around the world.At the same time,the global research on COVID-19 is also developing rapidly.Computerized tomography is an important method for physicians to detect and analyze the conditions of COVID-19 patients.How to use deep learning technology to process medical images and improve computer-aided diagnosis technology is an important research direction of deep learning technology in the field of medical image processing.A technique for deep learning to COVID-19 patients with CT image segmentation,can be a doctor in the clinical diagnosis and analysis provide effective technical support and auxiliary,thesis aimed lung CT image segmentation technology based on the deep learning,and CT image segmentation technology,based on the deep learning of new comprehensive analysis and the clinical application of COVID-19 disease,has the important clinical application value.The main research contents are as follows:1 An adaptive framework based on medical image segmentation technology is used.An adaptive framework of medical image segmentation technology based on the characteristics of medical image data was developed and deployed,and standardized pre-processing and segmentation training was realized for COVID-19 CT image data sets from different sources through the adaptive framework.2 A novel segmentation algorithm for COVID-19 lung CT scan based on Res-DC3 D U-Net neural network model was proposed.Res-DC 3D U-Net is a segmentation model based on deep convolutional neural network.Based on the structure of 3D U-Net,the convolutional module is replaced by the empty convolutional module(RES-DC block)with residual connection.The improved RES-DC-3D U-NET neural network,Combined with the expansion scheme of the data set,the simulation experiment shows that it performs well in the publicly released COVID-19 CT image verification set and has a leading segmentation level.3 A comprehensive analysis system of novel coronavirus CT image quantitative scoring based on deep learning was established.Based on the correlation between CT semi-quantitative score of lung involvement and clinical stage of the disease,an automatic analysis system for quantification of infected areas of COVID-19 image scan cases without manual intervention was designed,and the automatic analysis results were compared with the classification results of clinical cases.The experiment proved that: The automated analysis system can effectively simulate the doctor in clinical analysis,in-depth study of automatic scoring system overall accuracy reached 82.23%,for asymptomatic,mild,and normal phase of the forecast accurate rate was 80%,78% and 82%,with severe and critical(endangered)patient case to scan and prediction precision rate reached 89% and 100%,It can prove the effectiveness of the analysis system in the clinical rescue stage.4 Quantitative diagnosis and treatment based on deep learning technology is realized through intelligent segmentation of the lesion area and lung mask area,combined with the pathology and clinical practice of COVID-19,which can assist doctors to determine the stage of the patient’s disease and has practical clinical value for determining the diagnosis and treatment plan.