Detection of pulmonary tuberculosis with thoracic radiograph on ensemble deep learning model

The scarcity of abundant information for effective diagnosis and incomplete or unnoticed follow-up may trouble pulmonary tuberculosis patient. Any delay in diagnosing TB may lead to interrupting governments’ campaigns to eliminate national TB control programs. Improper planning and lengthy treatment may affect your bills. The wrong prediction may deteriorate an individual's health. All such cases need to be revised with an absolute and fast intelligent system of prediction of pulmonary Tuberculosis. Pulmonary Tuberculosis as a critical contagious issue requires opting for thoracic images to be applied with efficient Deep-learning, so a low-cost system can be developed for everyone in society. Through this paper, an effective deep learning model is presented to predict and classify pulmonary Tuberculosis among other pulmonary issues. The efficacy of the model relies on pre processing of dataset inputs. An accuracy of 96% is recorded on VGG16 whereas InceptionV3 resulted in 93% of overall accuracy.