Comprehensive Framework for Tuberculosis Detection Using Deep Learning and Image Processing in Whole-Slide Images

Tuberculosis (TB), caused by Mycobacterium tuberculosis (MTb), remains one of the leading causes of death globally, primarily affecting the lungs. Recent technological advancements have improved TB detection, particularly through the application of deep learning (DL) techniques. This research involves a pathologist using microscopy to identify acid-fast bacilli (AFB), combined with a proposed method that incorporates DL for TB detection. The DL models’ accuracy and efficiency are assessed and compared. The integration of DL aims to assist pathologists by providing an initial analysis, potentially streamlining their workflow. The proposed method provides the accuracy of 94%, which is equivalent to the traditional method. In contrast, the workload of pathologists, who is the bottleneck of the diagnosis, is considerably reduced. This novel approach not only demonstrates the potential of DL in healthcare but also emphasizes the significance of advanced diagnostic methods in addressing global health challenges like TB, ultimately improving healthcare outcomes and reducing mortality rates.