Bias in A.I. might be the biggest challenge ahead in making A.I.-based technology truly useful for medical purposes. But there’s a new paper in Nature Medicine that proposes to train algorithms to become better at predicting patient outcomes rather than to match the performance of their human counterparts. The paper presents this theory through a real-life example. Decades ago, a system was developed on white, British patients to determine their level of knee pain based on certain radiographic features of their X-ray. This system systematically scores black patients far below the severity of pain that they self-report. When they trained a deep-learning model to predict the patient’s self-reported pain level from their knee X-ray without using data from the old system, they found that the model was much more accurate, especially for black patients, thereby reducing the racial disparity. Learn more about A.I. in healthcare to be able to understand, anticipate and control it. Grab your copy of my e-book on The Medical Futurist ‘A Guide To Artificial Intelligence In Healthcare’: https://lnkd.in/gzEVGGN #themedicalfuturist #digitalhealth #future #healthcare #medicine #technology #AI #artificialintelligence #deeplearning #machinelearning …see more