A Review on Deep Learning-Based Segmentation Techniques for Lung Nodules

Lung cancer poses a serious threat to people’s health, thus escalating the necessity to have it detected and diagnosed in its early stages. Computer-aided diagnosis (CAD) systems are becoming a requirement for radiologists by reducing the time taken as well as the inter-observer variations to achieve this. Generally, the CAD systems for lung nodules include data collection followed by pre-processing, segmentation for pulmonary images, nodule detection, elimination of false-positive, segmentation, and then classification. This review focuses on the various techniques used for lung nodule segmentation. This segmentation is performed on computerized tomography (CT) images due to their high resolution levels. Furthermore, we discuss the various evaluation metrics and datasets involved in the segmentation of lung nodules.