Building Trustworthy Clinical AI: from Algorithms to Clinical Deployment
Jameel Clinic Collaborations Letter
The recent decade witnessed significant developments in clinical AI technologies. These advancements include all the areas of clinical care, ranging from risk assessment and diagnostics, to treatment personalization and prediction of treatment outcomes. If implemented in clinical practice, these new technologies can save lives and improve quality of care while controlling its costs.
However, translation of these technologies to hospitals has been lagging behind. For most of these powerful tools, we still need to understand how to best integrate them into the existing clinical pipelines to improve patient outcomes. This research has to be done in close collaboration with clinicians who will be ultimate users of the tools. Another important aspect of safe clinical AI deployment relates to its ability to robustly handle different patient populations, clinical settings, etc. We have already seen multiple cases where AI tools developed on one population were unable to scale up to other demographics, leading to inequitable outcomes in patient care. The only way to address this concern is by broadly testing AI tools across many hospitals that represent diverse demographics and clinical facilities.
A path forward for trusted AI in breast cancer risk prediction
As the role of artificial intelligence grows in medicine, one of the leading concerns is that algorithmic tools will perpetuate disparities...
Wellcome Trust Fellows