Pulmonary Tuberculosis Detection from Chest X-ray Using Deep Learning

Tuberculosis remains a formidable infectious disease, ranking among the top ten causes of global mortality. Timely detection is critical for effective treatment, yet current diagnostic methods face significant challenges. In this study, we propose a novel approach for automating tuberculosis detection from chest X-ray images. Our method integrates graph cut segmentation with convolutional neural network (CNN) classification, achieving an impressive accuracy of 94%, sensitivity of 96%, and specificity of 84%. This innovative approach holds promise for improving tuberculosis diagnosis, facilitating early intervention, and ultimately contributing to global tuberculosis control efforts. Key Words: Chest X-ray (CXR), Convolutional Neural Network (CNN), deep learning, graph cut, tuberculosis detection, automatic diagnosis.