Submit
This multicenter diagnostic study evaluated the accuracy of machine learning algorithms for tuberculosis detection using chest X-rays from 8,542 participants across 12 health centers in high-burden countries. Compared to microbiological reference standards, the deep learning model achieved 89.3% sensitivity and 92.7% specificity, outperforming both human readers and traditional diagnostic algorithms. Implementation in routine care could substantially improve TB case detection in resource-limited settings.
Keywords: Tuberculosis Diagnosis; Machine Learning; Diagnostic Accuracy; Global Health; Chest X-ray; Artificial Intelligence
