Tuberculosis (TB) Bacteria Detection Tool on Microscopic Images Based on You Only Look Once (YOLO)

Abstract

ABSTRACT Tuberculosis (TB) is an infectious disease caused by the bacterium Mycobacterium tuberculosis, which attacks the lungs and is one of the leading causes of death worldwide. This research aims to develop an automatic system for detecting tuberculosis bacteria from sputum microscopic images using the You Only Look Once (YOLO) v11 method, which is a branch of Deep Learning. The development process includes dataset collection and annotation, image preprocessing, model training, and system testing. The model is implemented using Python and OpenCV and is equipped with a Raspberry Pi 4-based GUI for easier user interaction. The detection results are visualized in the form of bounding boxes, accompanied by the number of bacteria and the severity of the infection (scanty, 1+, 2+, 3+). Based on testing 10 samples, the device successfully achieved 94% accuracy, and the system successfully reached a Precision value of 0.967 or 96.7%, Recall of 0.936 or 93.6%, mAP of 0.974 or 97.4%, and an F1 Score of 0.951 or 95.1%. These results indicate that the system is capable of effectively detecting TB bacteria with high accuracy.