Despite speech recognition (SR) technology seemingly becoming the next hottest trend with the introduction of Siri to the new iPhone 4S, it's still a not fully trustworthy technology in the realm of radiology, wrote Dr. Arun Krishnaraj in a recent commentary published in Diagnostic Imaging. Radiologists have used such technology for report dictation for years, but more often than not, are underwhelmed by the results, he said.
Specifically, Massachusetts General Hospital radiologist Krishnaraj cited two studies, one in which he participated, that reach conflicting conclusions about speech recognition software. A study in the American Journal of Roentgenology last year found that turnaround time for reports significantly improves with SR, and more frequent use results in smoother transitions.
However, a study published in last month's AJR tells a different tale. Reports generated via SR are eight times more likely to contain "major errors" than regularly transcribed reports, according to the study. What's more, Krishnaraj cited a Diagnostic Imaging poll that finds that 30 percent of responding docs who choose to use SR technology see it as inaccurate and difficult to use.
"[W]hat is good enough for the average consumer product doesn't necessarily stand up to the rigorous standards of healthcare," he wrote.
To learn more:
- read this commentary in Diagnostic Imaging