This systematic review and meta-analysis examined the rapid growth of machine learning techniques for chest X-ray (CXR) analysis in tuberculosis (TB) screening, particularly deep learning (DL) methods using convolutional neural networks (CNNs). Covering research published from 2016 to 2021, the review analyzed 54 peer-reviewed studies, identifying key datasets, methodological contributions, and challenges in TB detection. The findings highlighted significant advancements in DL techniques, including approaches that extend beyond binary classification to provide region-of-interest localization for improved diagnostic insights. While deep learning-based CAD systems show promise, the study underscores the need for standardized evaluation methods, diverse training datasets, and real-world validation to ensure robust and generalizable AI-based TB screening solutions.
