Continuous speech recognition in MR imaging reporting: advantages, disadvantages, and impact

Area of Science:

• Radiology
• Medical Informatics

Background:

• Conventional transcription of Magnetic Resonance (MR) imaging reports is time-consuming and costly.
• Advancements in speech recognition technology offer potential improvements in medical reporting.

Purpose of the Study:

• To evaluate a commercial continuous speech recognition (CSR) system for dictating MR imaging reports.
• To compare the efficiency, accuracy, and cost of CSR with conventional transcription.

Main Methods:

• A retrospective analysis of 5072 MR imaging reports generated using CSR was compared to 4552 reports from the previous year (conventional transcription).
• Data collected included report turnaround time, word recognition accuracy, report length, error rates, and system costs.

Main Results:

• CSR was used for 81.8% of reports, decreasing mean turnaround time from 87.8 to 43.6 hours.
• Report availability at 24 hours increased from 10.5% to 62.5% with 92.7% word recognition accuracy.
• Initial CSR costs were ~$10,000 versus $12,000 annually for human transcription, with a decrease in report length and spelling errors.

Conclusions:

• Continuous speech recognition markedly improved MR imaging report turnaround time.
• The system proved cost-effective compared to conventional transcription, despite being in early implementation stages.