Design And Implementation Of The Aided Detection System For Pulmonary Nodules Based On U-Net

Lung cancer is the cancer with the highest incidence both at home and abroad,and the mortality of lung cancer patients is also the highest among all cancers.Because patients with lung cancer have no obvious symptoms in the early stage,most of them have reached the advanced stage when they are diagnosed with lung cancer,and at this time,the surgical method can no longer play a good role,resulting in the low survival rate of patients.The five-year survival rate of lung cancer patients in stage Ⅰ or Ⅱ was 25%-73%,while that of patients in stage Ⅲ to Ⅳ was only 2%-24%.However,the early manifestation of lung cancer,that is,pulmonary nodules can be diagnosed by medical imaging,and pulmonary nodules can be detected earlier,which plays an important role in the early diagnosis of lung cancer.Based on the deep learning network model U-Net,this paper designs and implements a computer aided detection system for pulmonary nodules.The main research contents of this paper are as follows:(1)LUNA16 data set is used to train the U-Net model,and U-Net is used to segment the pulmonary nodules in the lung CT image data to realize the detection of pulmonary nodules.A variety of algorithms were used to compare with the U-Net network model,the experimental results were analyzed,and the existing problems in the U-Net were summarized.(2)analyze some existing problems of U-Net network and improve the network.The improved method is mainly as follows:the convolution layer is added into the skip connection structure of the original U-Net,so that the upper sampling path can obtain the high-level semantic information from the lower sampling path and the segmentation result is more accurate.Constraint conditions are added to the convolution layer to make the training speed of the network faster.(3)the pulmonary nodule detection system based on U-Net model was designed and implemented,which realized the detection function of pulmonary nodule in CT image data.In this paper,the existing lung nodules detection methods are analyzed and summarized,the use of U-Net network model to realize the detection of lung nodules in the lung CT image,aiming at the problems existing in the U-Net network model is improved and designed and developed based on U-net lung nodules aided detection system,realized the automatic detection of lung nodules,thus reduce the rate of missed diagnosis and misdiagnosis of lung nodules,reduce the doctor workload and improve automation degree of lung cancer detection.