Tuberculosis (TB) is a global threat to public health, and pathology is one of the most important diagnostic tools to treat it in clinical practice. However, due to the small size and the large number of bacilli bacteria, its treatment is time-consuming and requires an experienced pathologist. Thusly, in this age of AI, the development of modern computer techniques has escalated the process of diagnosing tuberculosis. In this article, Convolutional Neural Network (CNN) model has been developed to detect as well as classify tuberculosis in chest X-ray images. The model is trained using 700 TB and 3500 Normal images (publicly available images) taken from the Tuberculosis chest x-ray (CXR) dataset. Initially, the images are pre-processed by reducing their size and increasing their contrast to reduce noisy signals. Later, several parameters and layers are added to the CNN model for improving the prediction rate of tuberculosis detection. The applied proposed model achieved a 91.2% prediction rate with a minimal loss value, which shows a better prediction than the existing techniques applied on the CXR dataset.