<![CDATA[Tuberculosis (TB) is an infectious disease with a high mortality rate worldwide. Deep learning offers promising opportunities for automated TB diagnosis from chest X-ray (CXR) images. This systematic literature review (SLR), conducted following PRISMA 2020 guidelines, analyzes 66 articles from Scopus (2021–2026) to examine deep learning models, datasets, evaluation methods, challenges, and research opportunities. Findings reveal that CNN models remain dominant (42.42%), followed by hybrid CNN-Transformer models (37.88%), while public datasets are most frequently used (63.64%). Key challenges include dataset limitations, poor generalization, computational complexity, and lack of interpretability. This review contributes a comprehensive taxonomy of deep learning architectures for TB detection, identifies emerging trends toward hybrid and ensemble approaches, and provides actionable recommendations for future research, including federated learning, explainable AI, and clinical integration. These findings offer valuable guidance for researchers and practitioners developing reliable AI-based TB diagnostic systems.]]>
Copyrights © 2026
