Assessing and Mitigating Bias in Medical Artificial Intelligence
Background: Deep learning algorithms derived in homogeneous populations may be poorly generalizable and have the potential to reflect, perpetuate, and even exacerbate racial/ethnic disparities in health.
Deep learning algorithms using data from homogeneous populations may be difficult to generalise, and potentially exacerbate racial disparities in health and health care. This research paper explores (1) whether the accuracy of a deep learning algorithm varies depending on race / ethnicity, and (2) whether this performance is down to racial variation or external health factors.
