Thoracic X-ray features extraction using thresholding-based ROI template and PCA-based features selection for lung TB classification purposes

Abstract

This paper describes the results of research in finding the X-ray image features for the development of computer applications for identification of lung tuberculosis (TB) disease. We used statistical features of image histogram by calculates five features: mean, standar deviation (std), skewness, kurtosis, and entropy. These features were calculated from ROI images using pre-defined ROI shape from thresholding method. Average of trainer images was used in designing ROI shapes template using thresholding method. Features calculated was then reduced down to one principal feature using Principal Componen Analysis (PCA) method. This selected feature was to be used as descriptor in classifying image as TB or non-TB. We used Mahalanobis distance classifier to examined descriptor performance in image classification process. Image classification results show that features extraction can be done effectively using combination of thresholding-based ROI template and PCA (Principle Component Analysis) methods.