D.L.W.: The author is a consultant for Agfa Corporation.
Speech recognition has been used for a number of years for radiology reporting but its adoption remains controversial among radiologists. A number of studies have shown a time penalty in the use of this product. There is also concern over increased user distraction that may potentially diminish interpretation accuracy. Current systems are extremely user-dependent and a dedication to training and practice is often necessary for improved performance. Many radiologists are unaware of this and are using speech recognition in ways that have a negative effect on workflow and ultimately result in frustration with the product.A number of dictation techniques have been developed to maximize speech engine accuracy and consequently decrease correction time. A basic knowledge of the way the software functions in recognizing sounds can result in slight changes in dictation style resulting in improved accuracy and ultimate time savings. Microphone selection and positioning is an essential but often overlooked detail. Ongoing system training can be used to minimize the effect of a foreign accent. Rapid and efficient correction technique will also be reviewed.The use of macros and templates (stored reports) is common but not always utilized to full potential by radiologists. A simple and intuitive macro creation and naming scheme will be presented and demonstrated.Speech recognition systems accept voice commands for most navigational functions. Sometimes, however, other entry techniques are more desirable. The most common user interface, the keyboard, is often the worst device in terms of potential user distraction. The use of new hardware and software for "eyes-free" navigation through the speech recognition system will be discussed briefly.The integration of speech recognition with a Picture Archiving and Communication System (PACS) to run on a single platform is essential for efficient system use. A sample workflow algorithm for desktop integration of PACS and SR will be discussed briefly.A live product demonstration will follow with interactive problem-solving techniques discussed and presented.
To develop dictation techniques that maximize accuracy and navigation efficiency of speech recognition systems. Emphasis will be placed on minimizing the need for corrections, "eyes-free" navigation and SR/PACS integration software. A live demonstration of one speech recognition system will be performed using nontraditional navigation techniques.
Weiss, D,
Practical Methods to Minimize the Time Penalty and Distractions of Speech Recognition (SR) in Radiology Reporting. Radiological Society of North America 2004 Scientific Assembly and Annual Meeting, November 28 - December 3, 2004 ,Chicago IL.
http://archive.rsna.org/2004/4402850.html
