Segmentation and Visualization of Emphysema Lesions through Computed Tomography Images

Abstract

A large number of individuals in the world suffer from Chronic Obstructive Lung Disease (COPD). There is loss of the lung function in COPD due to emphysema and chronic bronchitis. The early detection of emphysematous lesions is very crucial using Computed Tomography (CT) images. This is intended to improve clinical management and prognosis. In this paper, an automated framework for robust segmentation of emphysema lesions is presented. It also presents morphological features using class specific intensity based thresholding, with morphological operations. The proposed method is based on the percentile-based threshold estimations adopted for each emphysema class. This is to facilitate the different density features and spatial distributions of these classes. The performance is evaluated on a multi-institutional dataset from Indian hospitals located in Chennai, Hyderabad and an online emphysema database. Quantitative analysis reveals distinct differences in the extent and distribution of lesions among the emphysema classes. CLE presents the maximum number of lesions, 61.49 per image and minimum burden (14.31%). PSE has the fewest number of lesions, 35.10 per image but high severity (19.76%). The 3D surface visualization helps to interpret the shape. It presents different intensity patterns and spatial characteristics for each of the emphysema classes. Segmentation and display in combination provide a comprehensive picture of the distribution and degree of emphysema. It serves as a useful tool for clinicians to enhance the diagnosis and determine treatment strategies.