Pulmonary Tuberculosis Bacilli Detection from Sputum Smear Microscopy Images Using K-Nearest Neighbor Classifier

Abstract

Abstract Tuberculosis is a very deadly disease worldwide, including in Ethiopia. TB is caused by Mycobacterium tuberculosis, which can cause pulmonary tuberculosis disease. Sputum smear microscopy is the most commonly used diagnostic tool in developing countries. The main purpose of this study is to develop a k-nearest neighbor classifier model for detecting PTB bacteria from sputum smear microscopic images. This study developed an algorithm based on the image processing technique to identify pulmonary tuberculosis bacilli in a digital image of a stained sputum smear. Thus, k-nearest neighbor classifiers were used to identify bacilli from sputum smear images in two classes: bacilli detect and non-bacilli detect. The total sample size of the image dataset of 180 from stained sputum images of PTB bacilli infected was obtained from Ethiopian Public Health Institute (EPHI). The model's accuracy, sensitivity, specificity, and F-measures then provided an average performance of 92.6%, which is the average performance of the prototype KNN model's sensitivity of 93%, specificity of 92%, and F-measure of 94.7%.