Diagnostic imaging is a popular clinical application of artificial intelligence and AI can successfully identify pattern changes that may be overlooked by human identification alone.  However, given that patient data from only three states trains most AI algorithms in the US, limited medical imaging datasets can result in diagnostic outcomes that are both geographic and demographic biased.

These companies offer products and, in some cases, solutions for rooting out bias early on in the development process and building diverse training datasets to prevent future inequities in healthcare.

---