Mycobacterium Tuberculosis Image Classifying Using YOLOv8 and Counting with IUATLD Assessment

Tuberculosis (TB) remains a significant global health challenge, with millions of new cases and deaths reported annually. Accurate and efficient detection of Mycobacterium tuberculosis is crucial for effective diagnosis and treatment. Traditional methods of TB detection are often time-consuming and require specialized expertise. This research aims to develop an automated system for classifying Mycobacterium tuberculosis images using the YOLOv8 model and to integrate the IUATLD assessment for accurate bacterial counting. Data collection used in this research includes image of Mycobacterium Tuberculosis from Department of Clinical Pathology Universitas Airlangga. The dataset that has been labeled in the previous stage will be trained to form a pattern whose results are in the form of weights. These weights will be used to detect objects in the image. Training is carried out using YOLO where counting tuberculosis used IUATLD Assessment. The device used in the training data retrieval process is Pycharm. Meanwhile, the YOLO used is YOLOv8. The Best result in data partition 60: 40 which produce an accuracy value of 63 %, a precision value of 69 %, a Recall value of 72 %, a mAP value of 84 % and a MAPE value of 0.76 %, whose interpretation of forecasting results is very accurate. The YOLOv8 model demonstrated high precision and recall rates, significantly improving the efficiency and accuracy of tuberculosis detection compared to traditional methods. The integration with IUATLD assessment provided a robust framework for quantifying bacterial load, which is crucial for diagnosis and treatment monitoring.