Analysis of local binary pattern for emphysema classification in lung CT image

Abstract

Medical digital images and methods for their processing and automatic analysis have been used for faster and more precise diagnosis. Computer-aided diagnosis systems are widely used by specialists as help for detecting and analyzing suspicions regions in medical digital images. Various types of medical digital images and numerous diseases that can be detected on them make this wide research field. One of the diseases that can be detected in lung CT images is chronic obstructive pulmonary disease or emphysema. In this paper we analyzed the capabilities of texture descriptors, local binary pattern, for detecting and classification of emphysema. Three different types of local binary pattern are used. Instead of using a whole local binary pattern operator output, statistical measurements have been used. Support vector machine optimized by elephant herding optimization algorithm was used for classification. Based on the obtained results, it can be concluded that six statistical information of uniform local binary pattern achieve the best classification accuracy.