Automated Detection of Lung Cancer using Transfer Learning based Deep Learning
Laxman Singh, Harsh Kumar Choudhary, Saurabh Singh, Ankit Kumar Bisht, Priyam Jain, Garima Shukla
Abstract
Lung cancer is a lung-affecting chronic disease that may severely affects the respiratory system. These days, lung cancer is known as the leading causes of death and is very difficult to cure at an advanced stage. Hence, detection and diagnosis of lung cancer at an early stage has become an essential requirement to offer a timely and successful treatment to the patient suffering from this disease. Recently, machine learning algorithms have gained momentum for the early diagnosis of disease that can assist specialist to greater extent in many ways. Hence, in the paper, the authors proposed a convolutional neural network (CNN) based on ResNet-50, a pre-trained CNN model, to uniquely classify the CT scans of lungs as cancerous or non-cancerous. The performance of the proposed system is assessed using 2478 lung CT scan images collected from LUNA 2016 dataset. The proposed transfer learning-based model achieved an accuracy of 99.1 per cent, which seems to be quite satisfactory in comparison to other recent state of art models reported in literature. Hence the proposed could be effectively utilized to assist doctors in interpreting and analyzing the lung cancer images for early diagnosis of lung cancer.