At last count, the Food and Drug Administration has approved 950 medical devices that use artificial intelligence and machine learning. But far fewer have had a significant impact on patient care in the United States.

This is a problem for medical AI developers like James Zou. The Stanford professor is investing in adopting effective medical machine learning techniques — including some of his own FDA-approved algorithms — into healthcare systems.

In two recent articles, Zou examined the clinical adoption of medical AI devices and found that a tiny minority of FDA-approved devices are reimbursed at some level – and examined different models for how health systems might pay for medical AI.