Feature Selection Based on Robustness for Classifying Lung Disease in Computer Tomography using Structural and Texture Feature

Computer aided system is not able to generalize the lung diseases . The main reason behind this is that in computer tomography, quantitative algorithm are not robust enough during image acquisition and reconstruction. It do not consider the technical factor. Thus algorithm is used that considers the robustness of the feature based on the technical parameter of the image. We evaluated various topographic images to classify various lung diseases. These images were separated into three datasets based on there texture and structural features. The classes of diseases that is considered includes idiopathic pulmonary fibrosis, interstitial pneumonia, cystic fibrosis, etiology, airways. Two classifier was created one with robustness feature and another without any robustness feature. Performance is compared on classifiers by using the mean, precision, recall, kappa and standard deviation