Detection of Tuberculosis from Chest X-Rays Using Convolutional Neural Network

Tuberculosis (TB) is a long-lasting respiratory disease that arises as a result of an infection of a bacterium called Mycobacterium tuberculosis. TB usually spreads when a person with active pulmonary and bacterium infection coughs, sneezes, or laughs in the presence of a crowd. TB diagnosis requires extraction of complicated TB symptom characteristics from chest X-rays, such as the presence of a lung cavity. Within this study, we present a machine learning architecture designed for robust detection using chest X-ray images as input. The concepts of image processing are used to define and identify features and CNN-based modeling is devised for deep learning classification, which is further improved by incorporating three different models that have been tried and compared. Transfer learning based on VGG19 architecture is also used to solve TB detection problems more effectively. Through the ML models deployed, the training and validation accuracies both reached around 95% within just 5 epochs. Both learning curves show promise that they could work on the VGG19 model for the detection of tuberculosis.