Tuberculosis (TB) is still a serious public health concern across the world, causing 1.4 million deaths each year. However, there has been a scarcity of radiological interpretation skills in many TB-infected locations, which may cause poor diagnosis rates and poor patient outcomes. A cost-effective and efficient automated technique might help screening evaluations in underprivileged countries and provide early illness diagnosis. In this work, we proposed a deep ensemble learning framework that integrates multisource data of two deep learning-based techniques for the automated diagnosis of TB. The integrated model framework has been tested on two publicly available datasets and one private dataset. While both proposed deep learning-based automated detection systems have shown high accuracy and specificity compared to state-of-the-art, the en- semble method significantly improved prediction accuracy in detecting chest radiographs with active pulmonary TB from a multi-ethnic patient cohort. Extensive experiments were used to validate the methodology, and the results were superior to previous approaches, showing the method’s practicality for application in the real world. By integrating supervised prediction and unsupervised representation, the ensemble method accu- rately classified TB with the area under the receiver operating characteristic (AUROC) up to 0.98 using chest radiography outperforming the other tested classifiers and achieving state- of-the-art. The methodology and findings provide a viable route for more accurate and quicker TB detection, especially in low and middle-income nations.