Cancer Stage Detection Using Deep Learning

Abstract

Lung cancer is one of the common types of cancer we see around the world. It’s very important to diagnose and start to treat lung cancer at an early stage, the earlier we start to diagnose it we will be able to save more and more lives dealing with it. There are different methods for diagnosing lung cancer including X rays, CT scans, PET-CT scans, bronchoscopies, and biopsies. However, staining techniques that stain tissue taken from a biopsy are often used to determine the subtypes of lung cancer based on H and E tissue, since knowing the sub type of lung cancer is very important, as it is curable if detected early. Histological examination of the suspicious tissue with regards to clinical and radiological features forms the definitive diagnosis of the disease. It is extremely important to analyse the histopathological image of lung cancer. Deep learning techniques are being used to speed up the critical process of lung cancer diagnosis and reduce the burden on pathologists and several studies have reported the importance of convolutional neural networks (CNNs) in classifying histopathological images of various types of cancer. Our objective is to provide a timely investigation to the people, a deep learning inceptionV3 model for early stage lung cancer detection is built using the dataset consisting of 5000 images of each :- Lung benign tissue, Lung adenocarcinoma , Lung squamous cell carcinoma. Model is built with an accuracy of 94%.