Deep learning to predict Pulmonary Tuberculosis from Chest Poster Anterior Radiographs of Lungs

On of the most serious health disease in all over the world is tuberculosis. Tuberculosis is an infectious disease as it affects the human body. According to the World Health Organization, approximately 1.7 million people affected with the tuberculosis during their lifetime. Pakistan is ranked fifth among high-burden countries and accounts for 61% of the tuberculosis burden in the WHO Eastern Mediterranean Region. Different methodology and techniques are available for the early detection of the tuberculosis, but these techniques and methodology have limitations. Majority of the techniques available in literature are using model-based segmentation of lung for the detection of tuberculosis. The main objective of the proposed research is to diagnose pulmonary tuberculosis using chest X- ray (Poster Anterior) lung images with a combination of image processing and machine learning techniques. The proposed research presents novel model segmentation approach for the detection of the tuberculosis. Different deep learning-based approaches are used for the classification like CNN, Google Net etc. The highest accuracy achieved for the proposed approach using Google Net as of 89.58% on combined datasets. The proposed research is helpful for accurate detection and diagnoses of tuberculosis.