Deep Learning Techniques for Discerning Various Phases in Lung Cancer Disease using Chest X-Ray Images

Abstract

Deep neural network models are boon to the society especially for medical field in diagnosing deadly diseases before any symptom could occur like cancerous cells. Usually predicting the cancer existence requires abundant time with skilled technicians to confirm whether cancer has invaded the human body. But the development of deep learning methods along with image processing relieved the entire process of detecting the infection earlier than other expertise. Cancer can affect any part in the body and lung cancer is predominant among other types around the world. So additional precaution is essential in diagnosing lung cancer in the premature stage. This paper focuses on the identification of lung cancer and various stages to determine the patient health condition using chest x-ray images. The images are acquired from diverse sources and preprocessed for further investigation. The images are then segmented using K means mask with bitwise AND operator and classified by deep learning methods to find the stages of cancer. The classification methods are Dense Net 121, Xception and Inception V3 model where best accuracy is produced by both models Dense net and Xception around 80% in classification.