Analysis of Recent Tuberculosis Detection Algorithms: Challenges and New Research Directions
Abstract
Nowadays, Machine Learning has become one of the important and emerging tool to classify and detect the infectious disease Tuberculosis. Unlike other infectious diseases, it is more difficult to diagnose Tuberculosis infection because several tests are required. Though the Chest X-ray (CXR) tool as a part of the World Health Organization is used to screen TB, the suspected individuals need different investigations before making the exact diagnosis and prescribing the necessary medications. Different emerging and attractive strategies such as Deep Learning, Neural Network and Machine Learning are designed for TB surveillance and detection. These strategies are working with Artificial Intelligence where the different interconnected components extract the required patterns from the real-world and complex input images. Recently these learning algorithms have proven the competitive results in a wide variety of sophisticated image-related operations in the radiology field. Since this medical domain primarily depends on the information extraction from real-world images, this has become a rapidly grown and active application research for many researchers. In this research, the opportunities, design, applications and limitations of very recent TB surveillance and detection algorithms on the different real-world datasets are analyzed in detail. This research also classifies the recent methods in some groups. This also analyzes the survey in terms of numerical insights and measurements of interests so that it will be very useful for the new researchers to develop and research further effective, improved methods in the right direction.