Published Paper Details:

This Tuberculosis (TB) is still a significant global health risk, especially in middle and low-income countries. WHO Global TB Report 2023 projected that about 10.6 million new cases of TB and 1.3 million deaths were reported in 2022 alone. Detection and correct diagnosis at an early stage are paramount to prevent the spread of TB and offer better patient outcomes. Chest X-ray (CXR) imaging is a routine TB screening because it is cost-effective with quick results. CXR is usually hampered by human factors, subjectivity, and the lack of trained radiologists, particularly in resource-poor environments. Recent developments in Deep Learning (DL) and Artificial Intelligence (AI) have raised hope towards computer-aided diagnosis of TB based on CXR images. In particular, Convolutional Neural Networks (CNNs) have shown superior performance in intricate feature extraction of features and feature classification accuracy in medical image processing. This review presents a critical overview of recent DL-based approaches to detection of TB with particular focus on CNN-based models. The paper also discusses the most significant challenges including data imbalance, image quality variation, problem of generalization, model interpretability, and deployment challenges in low-resource environments. Future research directions such as developing large-scale TB-specific datasets, building explainable and lightweight models, mobile health (mHealth) solutions, federated learning, and exploiting multi-modal data are highlighted. This article attempts to enlighten researchers, medical professionals, and policymakers about the current status, challenges, and future of DL-based TB diagnosis to guide international efforts toward efficacious and accountable AI-based health care solutions.