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

Lung cancer is a widespread disease,cancer deaths in many countries are caused by lung cancer,early diagnosis of this cancer can reduce unnecessary deaths.In the early stage of lung cancer,there are usually no obvious symptoms.It can only be found by computed tomography(CT),the early manifestations of lung cancer are usually solitary pulmonary nodules,so it is very important to detect and identify pulmonary nodules accurately.However,because of the huge amount of information on CT,it is necessary to detect small nodules with a diameter of only a few millimetres,which has great limitations only by naked eyes.Therefore,it is necessary to use computer-aided diagnosis(CAD)system to check and analyze patients’ lung CT images,so as to help doctors effectively improve the accuracy and efficiency of diagnosis of pulmonary nodules.As the most important part of CAD system,the classification of benign and malignant pulmonary nodules plays an important role in early detection and diagnosis of lung cancer.Convolutional neural network(CNN)has been widely used in the field of computer vision,so CNN has also been applied to the classification of benign and malignant pulmonary nodules.But there are some problems,such as low classification accuracy and high misdiagnosis rate.Therefore,this paper proposes an adaptive convolution neural network model and an improved convolution neural network model to improve the accuracy of classification of benign and malignant pulmonary nodules and reduce the misdiagnosis rate,thespecific research contents are as follows:The convolution operation of local receptive field in convolution layer of traditional convolution neural network is a single layer network,the inaccuracy of pulmonary nodule feature extraction results in low classification accuracy,in this paper,an adaptive convolution neural network model is constructed,the convolution layer of multilayer perceptron is used instead of convolution layer,which makes the neurons in each local receptive field perform more complex operations and extract more accurate pulmonary nodule features,and then improve the accuracy of classification.Then the influence of network depth and activation function on classification effect is analyzed,and experiments are carried out continuously to optimize the model and achieve good classification effect.In order to solve the problem of complex structure and too many parameters in the full connection layer of the adaptive convolution neural network,the convolution layer with fewer parameters is used instead of the full connection layer.Each feature graph is regarded as a feature of the output class,and the value of the feature graph is regarded as the confidence of a certain class,which is equivalent to the output eigenvector from the full connection layer,so that the feature map and category information are mapped directly.In order to solve the problem of parameter updating and learning rate selection,an effective parameter optimization algorithm and learning rate attenuation strategy are selected for the model through experiments,and then an improved convolution neural network model is constructed.Compared with the adaptive convolution neural network model,the training time has been shortened and the classification accuracy has been further improved.The experimental results show that the improved convolution neural network model constructed in this paper achieves good classification results,its accuracy,sensitivity,specificity and AUC values are 95.5%,0.96,0.95 and 0.96,respectively,this method has higher classification accuracy,lower misdiagnosis rate and missed diagnosis rate than traditional convolution neural network.Therefore,the research work of this paper has certain significance for the early diagnosis and treatment of lung cancer.