Pulmonary Tuberculosis

Deep convolutional neural networks (DCNNs) demonstrated high accuracy in classifying tuberculosis (TB) on chest radiographs, achieving an area under the curve (AUC) of 0.99. The study utilized both AlexNet and GoogLeNet models, with pretrained networks and data augmentation further enhancing performance. In cases of disagreement between classifiers, a radiologist’s review improved accuracy, resulting in a sensitivity of 97.3% and specificity of 100%.