CheXaid

A deep learning algorithm was developed to diagnose tuberculosis (TB) using clinical information and chest X-rays from 677 HIV-positive patients in South Africa. The algorithm, when used as a web-based diagnostic assistant, resulted in a modest improvement in clinician accuracy from 60% to 65%, though it maintained a higher accuracy of 79% on unseen test cases. These results suggest that the algorithm can enhance diagnostic accuracy in settings with high HIV/TB co-infection rates and may be particularly beneficial in environments with limited radiological expertise.