Texture Classi cation in Lung CT using Local Binary Patterns

In this paper we propose to use local binary patterns (LBP) as features in a classi cation framework for classifying di erent texture patterns in lung computed tomography. Image intensity is included by means of the joint LBP and intensity histogram, and classi cation is performed using the k nearest neighbor classi er with histogram similarity as distance measure. The proposed method is evaluated on a set of 168 regions of interest comprising normal tissue and di erent emphysema patterns, and compared to a lter bank based on Gaussian derivatives. The joint LBP and intensity histogram, achieving a classi cation accuracy of 95.2%, shows superior performance to using the common approach of taking moments of the lter response histograms as features, and slightly better performance than using the full lter response histograms instead. Classi cation results are better than some of those previously reported in the literature.