Tuberculosis TB Recognition System using Deep Learning Techniques

Abstract

Tuberculosis (TB) is a contagious disease which is one of the leading causes of death, globally. In India, more than 5000 people get infection every day and more than 1000 people die of TB, at the rate of 1 death for every 1.5 minutes. Since TB is communicable, its unpredictable growth rate of disease spread leads to high incidence and death rate. Microscopy is a rapid and early diagnostic method for many infectious diseases, including tuberculosis. In TB bacilli identification, specimens are stained using ZiehlNeelsen or Auramine dye and are thoroughly examined for finding out the presence of Acid Fast Bacilli (AFB) and its level of severity has been reported by skilled technicians. The entire slide viewing requires examination of approximately 6050 Field of Views (FOVs) per specimen, and a minimum of two samples are to be examined for every patient confirmation of TB. This process is laborious, involving fatigue, time consuming and delay in diagnosis, which leads to chronic TB infection in latent stage patients. Moreover, there are approximately 930 positively screened cases which does not fit in WHO severity grading and have intertwined severity levels. These cases have not been addressed/categorized in the World Health Organization s (WHO) severity level TB grading which causes inconsistency in diagnosis. Hence, there is a high demand for computer assisted system which could be used for early detection to assist pathologists with increased sensitivity and specificity. The proposed computer assisted TB detection system has three stages: Data acquisition system, Recognition system and Tuberculosis severity categorization stages. In the data acquisition system, a motorized microscopic stage is designed and developed to automate the acquisition of the entire eld of views. Here the microscopic stage movement is computer numeric controlled (CNC) and scanning patterns are given by the user for specimen examination.After the acquisition of all FOVs, data are passed to the recognition system.