Dr. Neil Shah, a radiology resident at NewYork-Presbyterian Hospital/Weill Cornell Medical College, will describe a test comparing consumer-grade speech recognition software (Dragon Dictation, Nuance Communications) with radiology speech recognition dictation software (PowerScribe, Nuance).
Ten radiology reports from exams ordered for emergency department patients were read and entered into an iPad. The same radiologist then read the report using a traditional workstation microphone and PowerScribe software. Each voice recording was transcribed into text and compared with the original report text.
The radiology-specific speech recognition system had an accuracy level of 97% based on the parameters defined by the researchers. By comparison, the iPad-generated transcription had an accuracy level of only 61% and required extensive corrections.