In CAD we trust

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon
 - Evan headshot 2013
Evan Godt, Editorial Director

Computer-aided diagnosis (CADx) and computer-aided detection (CADe) hold much promise for radiology. Effective use, though, remains a matter of trust.

That was the message from a review published in the February issue of Clinical Radiology. Wiard Jorritsma, a PhD candidate at the University of Gronigen in The Netherlands, and colleagues, noted that radiologists working as a sort of team with a CADx can produce effective reads.

“However, the team performance of radiologist and CAD is lower than what might be expected based on the performance of the radiologist and the CAD system in isolation,” wrote the authors.

The issue, it appears, may be trust. “Too little trust in a useful automated aid can lead to under-reliance, which means that the full potential of the aid is not being used,” Jorritsma and colleagues wrote.

There is a flip side to this dilemma, in that radiologists may trust CAD systems too much, which leads to over-reliance and errors that would not have been made by a human reader working independently.

CAD also can be costly. A study appearing last month in JAMA Internal Medicine that looked at CADe in mammography found that it’s used in nearly 80 percent of Medicare mammography exams. Over study period, which ran from 2001 to 2009, associated costs tallied nearly $300 million, and the researchers speculated on the cost of detecting ductal carcinoma in situ, which may be overdiagnosed.

Jorritsma and colleagues offered some suggestions to improve trust calibration between radiologist and CAD systems, including confidence ratings and more explicit rationales. You can read more of their suggestions here.

The potential for CAD is great, and more research will continue to define the sweet spot in which these systems offer the right amount of aid.

-Evan Godt
Editor – Health Imaging