Lung Cancer Classification Using Modified ResNet Architecture

One of the most prevalent and deadly cancers in the world, lung cancer presents serious difficulties for both patients and medical professionals. Even with the quick progress in early diagnosis, differentiating benign from malignant tumors is still a difficult task for doctors. By modifying the architecture and improving hyperparameters, this study aims to increase the model's accuracy in detecting subtypes of lung cancer and distinguishing between benign, malignant, and normal lesions. A new method for classifying lung cancer is provided by modifying the ResNet 18 architecture. While utilizing the basic hierarchical features that ResNet learnt, the suggested model incorporates modifications unique to the complexity of lung cancer imaging data. To assess the performance of the suggested model, the outcomes of this study (Modified ResNet 18) are compared with ResNet 18 and ResNet 50. Modified ResNet 18 exhibits the highest accuracy of 97%, followed by ResNet 18 at 95% and ResNet 50 at 95%.