PiXelNet: A DL-Based method for Diagnosing Lung Cancer using the Histopathological images

Abstract

Cancer is a group of diseases caused by abnormal cell growth, eventually leading to death. Cancer symptoms include chronic cough, breathing difficulties, weight loss, muscle stiffness, oedema, and bruises. Cancer detection increases with the stages, but unfortunately, the fatality also increases. In this research, we propose a pipeline coined as PiXelNet, which uses a classification system based on Convolutional Neural Networks (CNNs) that identifies three distinct kinds of lung cancer on histopathological images. The first step of the proposed network consists of a medical imaging analysis pipeline with models like ResNet, Efficient NetBO and MobileNet. We found that EfficientNet outperforms the other two models with a test accuracy of 99.33% and a loss of 0.0066. The second stage involves identifying the key areas from the original input test image with the feature extracted values. Using this strategy, the doctor or pathologist will immediately access all the crucial imaging heat maps and the network analysis report.