MedSpeak Gives Radiologists Advantage

Speech recognition technology developed by IBM and top radiologists is being used in a continuous speech dictation product that enables radiologist to work more efficiently while cutting health care costs.

MedSpeak/Radiology for Windows NT, with a vocabulary of 25,000 words has an average accuracy rate in beta tests of over 95% for most native North American English speakers.

The radiologist is the physcian responsible for interpreting X-rays, MRIs and CAT scans. Traditionally they have dictated patient information into a tape recorder for a transcriptionist who later typed up the report. That information was then returned to the radiologist for final review and signature, a process that can take several hours if the physician is able to attend to the matter immediately, and several days if he can not.

IBM’s MedSpeak/Radiology was developed with radiologists from two of the most renowned healthcare institutions in the country, Memorial Sloan-Kettering Cancer Center in New York and Massachusetts General Hospital in Boston.

It allows radiologists to dictate, edit and electronically sign their own reports in real time using a personal computer. This capability dramatically reduces report turnaround time and transcription costs and has the potential to increase the confidentiality of patient notes and charts through reduced reliance on outside services.

In the current managed care environment, faster results reporting speeds decision making, which in turn can shorten hospital stays.

In research studies conducted by the Massachusetts General Hospital, radiologists using MedSpeak virtually eliminated the administrative portion of the radiology dictation cycle.

“We have been following speech recognition at the Massachusetts General hospital for over 10 years, mostly with great disappointment, until our experience with the IBM MedSpeak/Radiology,” said Dr. James H. Thrall, radiologist-in-chief at the hospital. “From my standpoint of having responsibility for over 400,000 radiology reports each year, I see enormous potential for speech recognition technology to be a valuable and cost-effective tool for radiologists.”

Medspeak/Radiology is IBM’s premier solution offering for radiologists who need the productivity of continuous speech and integration with hospital information systems.

The MedSpeak solution consists of software and a special noise-canceling microphone, and runs on a personal computer requiring a Pentium Pro 200 megahertz processor. No special training of the software is required for most North American English speakers, although some users will benefit from completing a 12-15 minute enrollment in the system.

In addition, MedSpeak’s adaptive capability allows it to learn each radiologist’s dictation style as the software is being used. The advanced features and sophisticated language model of MedSpeak also allow the system to distinguish between words and symbols that sound alike, such as “to”, “two” and “too”; or the symbol for the punctuation mark colon (:) and a patient’s colon.

Performance can be further enhanced by the creation of personal or system wide “macros” which can generate entire sections of a radiology report at the utterance of a word.

“For those familiar with older voice recognition systems, MedSpeak/Radiology represents a quantum leap forward,” said Dr. Lawrence Schwartz, department of radiology. “The system allows users to speak completely naturally and accounts for differences in speech patterns, all with a very high accuracy rate. This facilitates reporting and eases the radiologists workflow.”

“Med Speak gives the radiologist a new, highly-advanced tool, one that will create a new way of handling the medical reporting process,” said Dan Bennewitz, general manager, North America, IBM Worldwide Healthcare Industry.

Flexible Configuration

The system can be configured in several ways, depending on the needs of the user. For a single practitioner, the system offers all of the functions of MedSpeak/Radiology on a single workstation, including permanent report storage and retrieval.

In larger practices or hospital settings, multiple MedSpeak workstations can be connected via a local-area-network (LAN) to a centralized server. In this configuration, any radiologist with an approved password can use any of the MedSpeak workstations with consistent recognition accuracy.

Preliminary reports can be centrally stored for editing by a transcriptist or even another radiologist.

Because Medspeak saves both the voice and text from the original dictation, a transcriber has the option of listening to the entire report, or only those sections or words which need correcting, saving time and keystrokes even in a traditional transcription environment.

The system can also handle multiple authors for each report, a plus in teaching hospitals where both residents and teaching staff are often involved in the same report.

MedSpeak can be linked to patient records on an existing hospital information system or radiology information system, creating a complete on-line record for the physician. The system contains a documented, HL-7 compliant message protocol that allows customers to download radiology orders (HSI) and (RIS).

IBM is creating and testing standardized interfaces with some of the industry’s leading vendors.

Contact Anthony Montesano at TSI for IBM at monty@tsipr.com.

First Products Using Dragon’s Engine Ready

Dragon Systems Inc., announced recently that the first products incorporting Dragon’s new large vocabulary continuous speech recognition engine will soon be available from the MRC Group for medical dictation/transcription.

MRC, a leader in medical transcription, working in collaboration with Articulate Systems, has introduced the ASI PowerScribe for Radiology system at the Healthcare Information Medical Systems Show in San Diego.

The first is a series of medical specialties, PowerScribe for Radiology, seamlessly integrates sophisticated automatic speech recognition transcription with standard digital dictation, providing maximum flexibility and utility for all users.

“After 15 years of creating successful technology tools and products leading up to this point, it is a pleasure for Dragon Systems to introduce its first commercial large vocabulary speech recognition systems, working with MRC and Articulate Systems,” said Janet Baker, president of Dragon Systems.

PowerScribe enables radiologists to use familiar handheld microphones with standard controls and a built-in speaker. Based on a client/server system configuration, PowerScribe allows physicians to move freely between different stations for dictating their reports. Over 26,000 medical vocabulary words are already built into the system.

For more information, contact Renee Blodgett at Dragon Systems, Inc. or renee@dragonsys.com