Lung cancer is one of the malignant tumors with high morbidity and mortality worldwide. Among them, non small cell lung cancer accounts for about 85% of all lung cancers. In the existing chest CT plain scan or enhanced scan, non-small cell lung cancer images may overlap significantly in imaging features, leading to misdiagnosis and the inability to give an accurate histological classification. Among the existing relevant non small cell lung cancer classification models, few studies on the precise classification of non-small cell lung cancer types. A multi-category classification model based on CNN is proposed to tackle these problems. The proposed model ISANET embeds channel attention and spatial attention mechanisms to focus on pathological areas based on InceptionV3. The three-category lung cancer dataset provided by the Affiliated Hospital of Hebei University is used, including lung squamous cell carcinoma, lung adenocarcinoma, and normal. Comparative experiments are done between ISANET and the traditional models AlexNet, VGG16, InceptionV3, MobilenetV2, and ResNet18. Results of experiments on two public datasets verify the effectiveness of ISANET, reaching 95.24% and 98.14% respectively, which indicates that ISANET has obtained superior accuracy in classifying non-small cell lung cancer.