Quantitative Analysis of HRCT Images for Characterization of Pulmonary Emphysema

This paper aims at improving quantitative measures of emphysema in computed tomography images of lungs. Pulmonary Emphysema is a chronic obstructive lung disease which is characterized by limitation of air flow. Detection and quantification of emphysema is important, since it is the main cause of shortness of breath and disability in Chronic Obstructive Pulmonary Disease. Current standard measures like Relative Area method, Pulmonary Function Test and Mean Lung Density methods rely on a single intensity threshold on individual pixels, ignoring any interrelations between pixels. HRCT is a sensitive method for diagnosing emphysema, assessing its severity, and determining its subtype. In this project texture based classification system is used. A study is performed by choosing Local binary Pattern (LBP) as texture feature and classification is performed using KNN classifier. Comparison is done with another texture feature i.e. Gaussian. The ROI classification using LBP showed good classification performance, compared to Gaussian. Thus LBP seems to perform better than Gaussian in finding the quantitative value. Also KNN have a greater sensitivity to emphysema. LBP analysis is a sensitive method for diagnosing emphysema, assessing its severity, and determining its subtype since both visual and quantitative HRCT assessment are closely correlated with the pathological extent of emphysema. Classification accuracy and quantification shows that KNN classifier performs better when LBP is used as texture feature in determining Pulmonary Emphysema.