Classification Of Benign And Malignant Pulmonary Nodules Based On Improved Convolutional Neural Network

The progress of the times is accompanied by the deterioration of the environment and the intensification of air pollution.Nowadays,malignant tumors,especially lung cancer,have become the first killer of the health that threatens people's lives.Early detection and diagnosis of lung cancer is of great benefit to improving patient survival.The early clinical manifestations of lung cancer are mostly isolated pulmonary nodules.The images show a round or round-like white area.If no measures are taken,it is quite difficult to judge by the naked eye.At the same time,due to the large number of CT image slices,it brings a great burden to the radiologist,which is likely to cause certain missed diagnosis and misdiagnosis.Therefore,this paper studies the image segmentation and nodule judgment in the diagnosis of CT images of pulmonary nodules.The main contents include:(1)This paper uses a method of segmentation of the region of interest in the lung.Firstly,the image enhancement is performed,and the histogram equalization method with contrast limitation is adopted to achieve the effect of not increasing the noise and improving the local contrast.Then,using the mean iterative method and the morphological method,the lung parenchyma is separated,and finally the morphological method is used to separate the ROI region..(2)Based on the original VGGNet network model structure,this paper improves the full connection layer of the last three layers,using the convolution layer instead of the full connection layer,and adopts dimensional reduction for the corresponding convolution layer.To increase the operation of nonlinear variation,an improved convolutional neural network model architecture is designed to maintain the accuracy of the model and accelerate the training speed of the network.(3)By comparing with the original network model and other classical models,the improved model adopted in this paper has reached a high level in accuracy,sensitivity and specificity,which are 90.51%,90.57% and 90.38%.