Superpixel and Density Based Region Segmentation Algorithm for Lung Nodule Detection

Abstract

Lung nodule detection and segmentation plays an important role in early cancer diagnosis. It is a challenging task owing to the shape and intensity variations of a lung nodule. This paper reports an efficient nodule detection framework in High- Resolution Computed Tomography (HRCT) images. Here, an automated computer-aided lung nodule detection scheme is proposed, combining the concept of superpixel generation and density-based region segmentation algorithm, Superpixel Density-Based Region segmentation (SPDBR). A set of morphological features are extracted from each of the extracted nodule regions. The nodule candidate regions have been classified into the nodule and non-nodule decision using a nonlinear support vector machine (SVM) classifier with an average detection accuracy of 84.75% with 82.86% sensitivity and 86.62% specificity.