Quantitative Analysis of Pulmonary Emphysema by Congregating Statistical Features

The chronic obstructive pulmonary disease (COPD) is a pulmo disease that is distinguished by reduction of peak expiratory flow rate and steady forced discharging of lungs. Pulmonary emphysema is one of the causes of COPD, and is due to damage in alveolar sacs that is caused due to loss in its elastic quality. An early diagnosis and it's treatment could help to take steps in controlling the symptoms. Hence, the quantification of emphysematous lesions in alveolar sac has been a significant topic of research. However, visual evaluation of computed tomography(Ct) images faces the challenge of being a tedious work and relies solely upon trained radiologist. Hence, an algorithm is proposed that would provide a reliable and automatic system that would end the dependence of radiologist and also aid in lessening time to accurately classify between normal tissue (NT), centrilobular (CL) and paraseptal (PS). The proposed novel method quantifies emphysema by congregating features: statistical features of wavelet transformed image, grey-level co-occurrence matrix(GLCM), first order texture statistical features and grey-level run length matrix(GLRLM). To validate the presented algorithm, leave-one-out cross validation technique was employed on the CT database. An accuracy and sensitivity of 91.67% and 91.89% was achieved. The specificity and precision of 95.72% and 92.22% respectively was obtained.