Segmentation of Tuberculosis Bacteria: Evaluation and Comparison

Tuberculosis (TB) is a communicable disease caused by Mycobacterium Tuberculosis that primarily affects the lungs. It is the second leading cause of death in developing countries and remains a major global health challenge in both developed and developing countries. In several countries in Africa and Asia, screening for pulmonary tuberculosis is done through Ziehl-Neelsen sputum smear images due to cost-effectiveness. Manual detection is time-consuming and strenuous work leading to misdiagnosis and low detection rate. It takes 40 minutes to 3 hours to analyze even 40 to 100 view field images from a single slide to diagnose a patient as tuberculosis positive or negative. This paper presents a comparative analysis of various supervised and unsupervised segmentation methods for automated TB bacilli segmentation in sputum smear images. The objective of this comparison study was to determine suitable segmentation techniques for sputum images. The quality of the images obtained from the smear depends on various factors, such as the staining procedure and the type of microscope used. Hence, it is essential to apply image preprocessing techniques to eliminate such variability by standardizing the RGB components of the image. The pre-processing techniques were analyzed in detail and an effective pre-processed image was determined using image quality metrics analysis. The results of the proposed study indicate that the UNet segmentation technique is far superior in terms of accuracy of 0.99.