HELPING RADIOLOGY

Dr. Wolfe says it might be possible to create exercises that could alter the prevalence rates, thereby helping agents at airport checkpoints find bombs, or radiologists in clinics find cancers, though at a possible cost of an inflated false alarm rate.


"We can move this behavior around in the lab," he says. For simple tasks, such as finding a red vertical line in a picture, subjects do better if shown the line beforehand. "If we give you a few minutes doing a 50 percent prevalence search, and give you lots of good feedback about how you're doing, when you go off to do this lab search in the lab, without good feedback [you're more likely to find it]" Dr. Wolfe says.

Dr. Wolfe's team is now testing the idea in real clinical settings. In experiments just beginning, he and his colleagues are slipping digital images into radiologists' workflows to see if their ability to spot lesions is affected by how commonly seen they are.

Still, for radiologists, it's possible that a slight false positive bias - driven by worries about missing potentially deadly cancers and malpractice fears - could drown any effect seen by toggling with background rates. Also, "signal detection" tasks in radiology are rarely simple binary operations, with yes or no answers, and could be harder to prime radiologists for.

Dr. Wolfe hopes to have some answers soon. "Check back in a year," he says. "We'll see."

Back to HCB News

More Industry Headlines