Research On The Detection And Extraction Methods Of Lung Nodules Based On CT Images

In 2018,the official journal of the American Cancer Society,Journal of Clinician Cancer,announced that the incidence and mortality of lung cancer in the world ranks first among all malignant tumors,which seriously threatens people's lives.The survival rate of lung cancer is closely related to the disease stage when the patient is first diagnosed.Because the symptoms of early lung cancer are mostly inconspicuous,the disease stage often reaches the middle and late stage when clinical diagnosis is made.Therefore,it is very important for early accurate detection and diagnosis of lung cancer During the research,it was found that the manifestations of lung cancer were mostly pulmonary nodule lesions.This study is designed to detect and extract lung nodules based on artificial intelligence algorithm for CT scan images of lungs to assist physicians in the task of accurate detection and diagnosis of pulmonary nodules:(1)A pulmonary nodule model was constructed by performing Deep Convolutional Neural Network(DCNN)training on lung CT images of lung nodules;The trained lung nodule model is used to predict the data to be measured,and the location of the pulmonary nodule and the probability value of the pulmonary nodule are obtained.Among them,DCNN training lung nodule model algorithm includes:convolutional neural network extracts the feature map of lung nodule image,RPN network extracts lung nodule candidate region and region of interest(ROI)classification and regression.Thereby the probability value of the lung nodule and the position coordinate value of the lung nodule are obtained.It solves the problem that there are many types of pulmonary nodules in CT scan images,and the difference is large,which makes it difficult to achieve high accurate detection and localization.(2)Based on accurate positioning of lung nodule regions based on machine learning,the lung morphology images were obtained by mathematical morphology processing of the lung CT images;Obtaining an initial contour image of the lung nodule with the background according to the detected coordinate position of the lung nodule;Accurate segmentation extraction of lung nodule contours by utilizing an Expectation Maximization(EM)algorithm,that is,calculating an expected value E step and maximizing M steps to remove the background;Calculate the key feature information of lung nodules by mathematical statistics method,including lung nodule volume,area,gray mean value,gray scale variance,skewness value,kurtosis value,etc.These features help doctors diagnose the benign and malignant lung nodules and provide better treatment options for patients.(3)In order to truly realize the assistant doctor's task of detecting and diagnosing pulmonary nodules,the lung nodule detection location,segmentation extraction and key feature value information calculation are integrated,and the pulmonary nodule visualization auxiliary diagnosis system is developed.The doctor can operate interactively.The system interface is more direct and convenient to make accurate diagnosis of lung nodule pathology,saving manpower and time cost of hospital pathology,improving the quality and efficiency of pathological diagnosis,and thus has great significance for driving the development of precision medicine.