MULTILEVEL DEEP LEARNING TECHNIQUE BASED CLASSIFICATION OF LUNG CANCER AS PER TUMOR NODE METASTASIS CODING

Abstract

A multilevel deep learning approach known as TNM coding is commonly employed to classify lung cancer and predict disease stages. This method integrates multiple deep learning networks to address complex challenges effectively. The objective of this study is to categorize CT lung scans into three Tumor-Node-Metastasis (TNM) staging classes in accord with the American Joint Committee on Cancer's (AJCC) guidelines. Initially, different lung diseases such as juxtapleural and internal nodules are segmented automatically using an optimized conditional generative adversarial network (c-GAN). Following that, combination of support vector machine classifiers and deep learning models are used to categorize tumors, nodes, and metastases based on the AJCC staging criteria. This automated TNM cataloging method provides accurate cancer staging without requiring manual detection of the region of interest (ROI) within CT scans. Several conventional approaches classify nodules into one of two categories—cancer versus non-cancer—whereas the proposed technique offers a holistic classification model. It improves upon existing binary classifiers and achieves results comparable to sophisticated TNM staging systems. This method enhances the precision and cost-effectiveness of CT scan analysis for lung cancer classification by streamlining the categorization process and reducing dependence on human expertise.