Four things to know:
1. The technology is aimed at completing routine tasks for physicians. Pranav Rajpurkar, a computer science graduate student at Stanford (Calif.) University, in 2017 developed an algorithm that used chest X-ray data from the National Institutes of Health to teach itself how to diagnose conditions from the scans.
2. While the algorithm had no idea what to specifically look for, it taught itself over time by searching for patterns, a technique known as deep learning. Today, the same black box model is being used to diagnose tuberculosis in HIV-positive patients in South Africa.
3. The algorithm correctly diagnosed TB 75 percent of the time. While not perfect, this is an improvement from the 62 percent rate of physicians in South Africa. Stanford radiologist Matthew Lungren, MD, is the main medical adviser for the project. He similarly misdiagnosed his first two cases due to TB not being common in the Silicon Valley, he said.
“The ultimate thought from our group is that if we can combine the best of what humans offer in their diagnostic work and the best of what these models can offer, I think you’re going to have a better level of healthcare for everyone,” Dr. Lungren told NPR.
4. However, these algorithms do have their downfalls. A medical resident at California Pacific Medical Center in San Francisco found a loophole in Mr. Rajpurkar’s first algorithm that used X-ray scans to diagnose patients. He said the algorithm “cheated” by considering other elements of the scan, such as information around the edges that indicated the type of machine used to make diagnoses.
To read the full report, click here.
More articles on health IT:
Health tech, subscription services will drive Apple’s profits, former CEO John Sculley says
Apple extends iPhone health records access to Allina Health
Wage lawsuit against Epic tossed by judge & 8 other health IT notes