Automated Process of Identifying Tuberculosis in Ziehl Neelsen Stain

Abstract

Mycobacterium Tuberculosis is a serious illness for which early diagnosis is critical for disease control. Common method of TB diagnostic is visual analysis of sputum samples seen under a microscope. This manual process is time consuming, subject to poor speci- city, labor intensive and a tedious task. A prototype system was developed using image processing technique to get rid of the drawbacks of the manual searching process. Here we present a method of automating TB diagnostic process in Ziehl Neelsen staining using image processing techniques. Our algorithm employs color space techniques, morphological operations, size ltering and color ltering methods for segmentation process. We propose a novel algorithm to bacilli classi cation and counting based on area and perimeter shape descriptors. These classi cation technique was applied for counting purposes and the sensitivity vs specicity results were evaluated using a standard ROC analysis procedure.Experimental results con rmed the superior performance of the proposed algorithm.