Lung cancer is a high incidence disease, which seriously affects people’s health. The pathological section of lung cancer can determine the type and differentiation of lung cancer cells, so as to provide an important basis for the selecting treatment options. In recent years, researchers focus on using convolutional neural network (CNN) algorithm to assist doctors and improve the recognition of cancerous regions in pathological images. In this paper, the CNN models were used to identify the cancerous region of lung cancer pathological images. The public data set was selected to train the AlexNet, GoogLeNet and ResNet34 models, and adjust the relevant parameters to improve the accuracy and specificity of recognition as much as possible. The experimental results showed that the accuracy and specificity of ResNet34 model were 98.9% and 99.0%, respectively, indicating that the model could effectively assist doctors to identify cancerous regions in lung cancer pathological images.