Abstract
Tuberculosis (TB) is a serious infectious disease that mainly affects the human lungs. The bacteria that cause TB are spread via minute droplets released into the air via sneezes and/or coughs. A bacterium called Mycobacterium is the root cause of TB. This paper is to investigate the precision of four factors of detecting Pulmonary Tuberculosis based on the patients’ chest X-ray images (CXR) using Convolutional Neural Networks (CNN). We evaluate image dataset resolution, and then the pre-trained networks (AlexNet, VGG16 and VGG19) and various hyperparameter changes are investigated. Finally, additional sample images are tested and investigated. Simulations have been carried out based on 406 normal images & 394 abnormal images. Later an additional 239 normal images and 554 abnormal images are added. It is found that the splitting of images yielded the best results.