A STUDY ON LUNG CANCER: DETECTION AND CLASSIFICATION

Abstract

This paper covers lung tumor segmentation using two proposed networks. The first is the 3D densely connected convolutional neural network. And the second is a 3D convolutional neural network with deformable convolutions. The working of all networks is presented in the paper, along with their qualitative and quantitative results. Moreover, the lung tumor datasets used for the proposed studies are discussed in the paper. A comparative analysis of lung tumor segmentation using different networks is presented. Lung cancer is primarily diagnosed in later stages when cancer has spread to a much larger extent resulting in big and irregular-sized tumors. It is required to segment the lung tumor to analyze it so that doctors can decide on the appropriate treatment planning. We proposed three approaches for lung tumor segmentation which are discussed in this paper.