Revolutionizing Tuberculosis Detection: Advanced TBNet Classifier for Chest X-Rays

Despite marked improvements in treatment over the decades, tuberculosis (TB) remains a prominent global health issue burdening physicians and society alike, and characterized by a need for accurate and rapid diagnosis methods to control its geographical spread and increase patient efficacy. In this study, we introduce Advanced TBNet Classifier, a deep learning-based model that reliably and very accurately categorizes tuberculosis images. The model uses robust CNN architectures to extract in-formative features with the support of preprocessing methods to attain better image quality for possible better model performance. This work leverages LIME to ensure openness in the diagnostic process through visualizations of interpretable components of the input images explaining model predictions. The interpretability provides healthcare providers insight into the decision-making process, which in turn builds trust. The Advanced TBNet Clas-sifier demonstrates exceptional performance measures with an accuracy, precision, recall and, f1-score that highly aligned with the contribution to this study and is of 99.57%, along with AUC of 99.99%. The accuracy of these findings as a replaceable TB detection tool for doctors in resource-constrained environments is highlighted and diagnostic delays are reduced.