ABSTRACT
This article proposes a novel technique to detect lung cancer from CT images that use deep residual learning. To identify areas of the lung that are at risk for cancer, we outline a series of pre-processing approaches. After the data is analyzed using classifiers such as XGBoost and Random Forest, it is integrated with other information to determine whether or not a CT scan is cancerous. 3D-SE-IRNet and 3D-SE-IRNet were created to increase network accuracy by making use of the 3D nature of CT scans and feature map channels. The test we did shows that CNNs are good at putting lung tumors on CT pictures into the right category. Self-mechanism and 3D convolution make the network much more accurate. Specificity, and sensitivity. This shows that the individual model is not as good as the Ensemble model. In addition to this, the E-ResNet-NRC model has improved generalization and resistance capabilities.
