Lung Disease Differentiation and Cancer Staging Using CNN

Abstract

World Health Organization (WHO) acknowledged COVID-19, Cancer, and Tuberculosis as global epidemics in March 2020. Machine learning (ML) techniques can play crucial roles in the detection and identification of these respiratory diseases. This paper introduces a machine learning approach aimed at categorizing X-ray images of chest into four distinct classes: COVID-19, normal, lung cancer, Tuberculosis. The architecture network used for this classification is VGG16 Convolutional Neural Network (CNN. The accuracy for grouping achieved through this is 97%. In addition to disease differentiation, lung cancer staging is attempted using CT scan images. Three levels of staging are identified names as: early-stage, locally advanced, and advanced-stage, which will aid clinicians in making informed treatment decisions. Here two separate models using CNN and Random Forest are implemented and compared for staging. Radom Forest classifier gives 83% accuracy whereas CNN gives an accuracy of 98%.