Segmentation of Pulmonary Parenchyma in Thoracic CT Images Based on 2D Otsu Optimized by DPSO

Abstract

Lung segmentation is considered as the basic foundation for computer aided diagnosis (CAD) of lung diagnosis. In this paper, two-dimensional Otsu method based on the Darwinian particle Swarm Optimization (DPSO) is proposed to segment pulmonary parenchyma from lung Computer Tomography (CT) scan images. The CT scan and CAD facilitate researchers to compel cutting edge image processing techniques for identifying lung cancer and probable other lung illness including emphysema and bronchitis. The 2D Otsu algorithm is used for CT lung image segmentation which plays a vital role due to better performance results even in the noisy environment. The main drawback in this algorithm involves computation complexity and time. In this paper, modified and optimized 2D Otsu by DPSO is developed to diminish the drawback of conventional 2D Otsu method. In this work DPSO algorithm is used to get better optimal threshold for segment pulmonary parenchyma in less computing time. Test results demonstrate that the proposed technique gives segmentation results similar to 2D Otsu method but improoved results, when considering different measures. The algorithms tested on the Cancer Genome Atlas-Lung Squamous Cell Carcinoma (TCGA-LUSC) datasets.