Computeraided Diagnosis Of Lung Tumors Based On Deep Learning

At present,the incidence rate and mortality rate of lung cancer are increasing every year.Because of the lack of obvious symptoms in early stage of lung cancer,patients are often not easily found.When the patient has obvious symptoms,the cancer cells have spread and reached the stage of advanced lung cancer.With the development of artificial intelligence,deep learning has made remarkable achievements in medical image processing,and medical intellectualization has become a hot research field.Firstly,the lung CT image data set is preprocessed,and the lung parenchyma is segmented by threshold segmentation method and mathematical morphology method.The segmented lung parenchyma image is used for the detection of pulmonary nodules and the screening of false positive pulmonary nodules.In the stage of pulmonary nodule detection,yolov3 network is selected by comparing the mainstream target detection algorithms and combining with the characteristics of pulmonary nodules.The average accuracy of the improved network is higher than that of the original yolov3.In order to avoid missed diagnosis,suspected nodules are extracted as much as possible in the detection stage of pulmonary nodules,which leads to the phenomenon of false positive nodules in candidate nodules.Through data enhancement processing and calculating the loss value of positive and negative samples separately,the phenomenon that the loss value is dominated by the original imbalance is reduced.Through the training of the model,higher specificity and sensitivity value were obtained,which showed that the model has strong filtering ability for false positive nodules,and the possibility of missed diagnosis is also small.Figure 36;Table 5;Reference 53...