Project Intro
Tuberculosis (TB) is an airborne disease caused by Mycobacterium tuberculosis (M. tuberculosis). It is a global infectious disease problem, especially prevalent in undeveloped and developing countries. According to the WHO 2017 report, it is one of the top ten causes of death worldwide and currently the most contagious disease. Conventional light microscopy is the most widely used detection method. Pathologists check for the presence of M. tuberculosis through the acid-fast stained sputum smear. The disadvantage of this method is that the process requires a lot of manpower, and the accuracy is only about 50% to 70%. Therefore, for disease prevention, it would be very beneficial to have an automatic, fast and accurate M. tuberculosis identification system. In this project, artificial intelligence (AI) deep learning technology is applied to develop an automated M. tuberculosis identification system. Simultaneously, deep learning optimization methods will also be improved and optimized in order for medical exam speed and accuracy of TB to be effectively increased. Academic contribution will come from the innovation of the deep learning optimization improvements in the study. Relevant results can then be extended to the identification of nontuberculous Mycobacteriosis (NTM) and extrapulmonary TB. All of the aforementioned can provide thorough assistance for tuberculosis control. Not only does this project provide a great extent of AI technical education, the results could also easily be transferred to the industry, which may ultimately increase the competitiveness of AI medical devices of Taiwan.