Grant: Project Grant

Although artificial intelligence (AI) is now widely being used to improve health, there are challenges in harnessing innovation for those most marginalized. Workers in South Africa's gold mines have had the world's highest tuberculosis (TB) rates, as exposure to silica dust increases TB vulnerability - even in people who do not have silicosis, itself often a debilitating disease. Over half of this workforce are migrants; and when they return home with TB, they spread disease to their communities. Efficient detection of TB and silicosis is especially needed to provide afflicted workers and their families with needed social benefits and health services to which they are entitled. While all parties, including government authorities and mining companies as well as workers and their representatives are now committed to addressing this problem, there are not enough trained health professionals to confidently conduct health assessments in areas where most ex-miners live. While Computer Assisted Detection (CAD) systems having been developed to detect TB in chest x-rays, none of these systems have been implemented and validated for people who also may have silicosis, a disease that can mimic and mask TB. Our research will adapt recent innovations in machine learning to demonstrate the feasibility of overcoming this equity challenge in field tests in diverse sites (MBOD, 2 mining company clinics, 2 mobile clinics including 1 outside South Africa, and a hospital site). Our interdisciplinary team has a strong track record for conducting this research, with experts in state-of-the-art AI advances, radiology, lung disease and social sciences, including gender and other equity issues. Interviews and workshops with those directly affected by implementation of this innovation will identify ways to overcome social, political, economic, legal, gendered and ethical barriers to applying this equity-sensitive decision-support tool to benefit current and former miners and their families.