A Morphological Approach for the Extraction of the Disorders of the Pleura from Chest CT Slices Anita Titus 1 , H. Khanna Nehemiah 2,∗ , A Kannan 3 and J. Jabez Christopher 4 1,2,4 Ramanujan Computing Centre, 3 Department of Information Science and Technology, Anna University, Tamil Nadu, Chennai 600 025, India. e-mail: nehemiah@annauniv.edu Abstract. A Computer Aided Diagnosis (CAD) system is proposed for the detection of pleural effusion and pneumothorax, which affect the pleural membranes of the lungs. The accuracy of a CAD system is largely dependent on the efficiency and preciseness of its segmentation system. The segmentation technique used here extracts the lungs and the regions affected by pleural effusion using conventional thresholding techniques like Otsu’s and iterative thresholding, followed by morphological operations. The chest CT slices are initially preprocessed to remove the gaussian noise by using a Gaussian filter. Morphological operations are then applied, to segment the lung parenchyma as well as to extract the region of interest (ROI). Texture features are next extracted from the ROIs, which are used to compute the feature vectors. These are used to train a neural network classifier. After the training process is complete, the query images are given to the system for classification. The CAD system achieved an accuracy of 95.458%, sensitivity of 88.92% and specificity of 97.918%. Keywords: Computer aided diagnosis (CAD), Lung segmentation, Pleural effusion, Pneumothorax, Morphological operations, Probabilistic neural network, Classification. 1. Introduction The lungs are surrounded by two serous membranes, the parietal pleura and the visceral pleura. The parietal pleura is the outer membrane that is attached to the wall of the thoracic cavity and the visceral pleura is the inner layer of the membrane that covers the outer surface of the lungs. The pleural cavity refers to the space between the parietal and visceral pleura. A small amount of fluid is secreted to maintain lubrication between the two pleural membranes. When there is excessive accumulation of the fluid in the pleural cavity, it results in a pleural disorder called as pleural effusion. Pneumothorax is a pleural disorder where there is excessive collection of air in the pleural space which when severe causes the entire lung to collapse. The presence of pleural effusion or pneumothorax in the pleural cavity results in the lungs getting compressed. This in turn results in lesser amount of oxygen entering the lungs. Pleural effusion occurs in 15–20% of patients affected by Tuberculosis (TB) and may be present on the same side where the TB is located [1]. The lung parenchyma is porous and spongy in nature, consisting of minute air sacs called alveoli which help in the exchange of gases. The pleural fluid accumulates at the lower portions of the lungs as the density of the pleural fluid is greater than the density of the lung parenchyma. The pleural fluid takes the shape of the lungs and the pleural cavity. Pneumothorax is a pleural disorder resulting in the accumulation of air in the pleural space. As air is less dense compared to the lung parenchyma, the pneumothorax region will also take the shape of the lungs and the lung cavity and will commonly occupy the upper regions of the lungs. The grayscale value of air corresponds to values close to zero (black) and that of water or fluid is close to 128 [2]. Hence the entire pleural effusion region in the CT slice has a grayscale value close to 128. Similarly the entire pneumothorax region in the CT slice has the same grayscale value as air, which is close to zero (black). Most of the segmentation schemes in the literature have focussed on the extraction of the lung parenchyma, so a technique is proposed for the segmentation and extraction of pleural effusion and pneumothorax, both of which are present in the pleural region outside the lung parenchyma. ∗ Corresponding author  Elsevier Publications 2014. 699