Tumour Node Metastasis (TNM) coding is used to classify lung cancer using a multilevel deep learning approach, which aids in predicting the stage of the disease. Using many deep learning networks at different levels, the multilevel deep learning approach handles complicated issues. In order to classify CT lung scans into three groups of classes, this study developed three unique Tumor-Node-Metastasis categorization methods, as advised by the American Joint Committee on Cancer (AJCC). First, an ideal conditional generative adversarial network (c-GAN) has been established for automated lung segmentation, which includes the juxtapleural nodule and the nodules inside the lung. For the next step, three support vector machine (SVM) classifiers and three distinct deep learning algorithms have been created for cataloging of tumor (T), node (N), and metastasis (M) according to AJCC staging terminology. This learning offers the TNM cataloging, which aids in diagnosis of the cancer stage, in contrast to previous studies on lung cancer classification that concentrated on classifying the given nodule as cancerous or non-cancerous. The suggested classifier outperforms the currently used binary cancer classification algorithm and achieves performance that is comparable to that of the recently created TNM classifier. The suggested method also has the benefit of not requiring human skilled participation to identify the ROI area of the image in order to classify given CT picture into TNM classes.