Pulmonary Tuberculosis Classification on Chest X-Ray Images using an Attention-Based Convolutional Neural Network (CNN) and Bayesian Hyperparameter Optimization

Pulmonary tuberculosis (TB) is a global health problem that requires accurate early detection. TB detection is generally performed through the interpretation of chest X-ray images by medical personnel, but this is subjective and dependent on individual expertise, which can lead to delays and inaccuracies in diagnosis. Advances in artificial intelligence technology, intensive learning, offer automated solutions for medical image analysis. This study proposes a Convolutional Neural Network (CNN)based classification model equipped with an attention mechanism to increase focus on medically relevant image areas. Furthermore, Bayesian-based hyperparameter optimization is applied to achieve optimal model performance. The proposed model achieved 94.28 % accuracy, 92.85% precision, 95.71% recall, and 96.30% AUC, demonstrating its effectiveness in supporting accurate and consistent TB detection from chest X-ray images.