Lung Cancer Classification and Prediction of Disease Severity Score Using Deep Learning

Abstract

The World Health organization (WHO) recent statistics show that Cancer is a life-threatening disease that causes 10 million deaths every year around the globe. Lung Cancer is a leading cause of death worldwide, accounting for nearly 2.21 million deaths in 2020. Lung cancer is increasing day by day so early detection is much needed to initiate proper treatment to save the life of cancer patients. Lung cancer detection at an early stage has become very important and easy with image processing and deep learning techniques. The proposed work uses histopathological images (microscopic examination of a biopsy) to classify different cancer categories. This paper presents the use of Adaptive fine-tuned EfficientNetB7 architecture to classify three categories (2-cancer types Adenocarcinoma, Squamous cell carcinoma, and 1-normal i.e benign). The classification results enable the doctors to detect benign or malignant categories to initiate proper treatment. In this work measured performance ma such as Recall, Fl-Score, Precision, and classification accuracy. The proposed work enhanced the classification accuracy from 97.5% to 99.5% compared to the existing work. Later predicted the disease severity score in four levels based on the number of diseased cells present in the image.