NYU plans to use AI to check CT scans for osteoporosis

Katie Palmer covers telehealth, clinical artificial intelligence, and the health data economy — with an emphasis on the impacts of digital health care for patients, providers, and businesses. You can reach Katie on Signal at palmer.01.

If a patient goes in for an X-ray or CT scan these days, that image is increasingly likely to be analyzed by artificial intelligence. But a growing category of algorithms is being used not just to help radiologists interpret images, but to catch hidden signals within them. 

The Food and Drug Administration has cleared several AI algorithms to flag “incidental” findings like pulmonary embolisms, calcifications in arteries, and low bone density in scans captured for other reasons. Together, they signal a future in which radiologists can make more out of the mounds of images they’ve already collected — and possibly identify problems before they become life-threatening.

This kind of opportunistic screening has been a goal in radiology for decades but is rarely implemented at scale. “As long as it’s not there with one click going into the report, it’s hard to really do it universally,” said Miriam Bredella, vice chair for strategy at NYU Langone’s radiology department. “And you only do it well if you do it in everyone — not just in people you’re worried about — because you want to detect the ones that would have fallen through the cracks.”