Comparative Analysis of Lung Segmentation

Abstract

Lung cancer detection is a crucial and time-consuming task which is undertaken majorly by highly trained professionals or doctors. Lung parenchyma segmentation is a pre-processing technique for lung nodule detection. Segmentation of lung parenchyma is useful for delineation of lung nodules or lesions and for further analysis. It can be achieved by automating the segmentation process to reduce human errors as well as to help radiologist for accurate diagnosis of lung diseases. This paper deals with comparative analysis of three lung parenchyma segmentation techniques. Watershed segmentation provided a better accuracy as compared to region based segmentation and cluster based segmentation. The segmentation algorithms are tested on 400,000 images. Watershed segmentation outperforms and provides an accuracy of 96.57%.

Keywords: Lung cancer, Lung parenchyma, Lung nodule detection, Cluster based segmentation, Region based segmentation, Watershed Segmentation

JEL Classification: Y60