PACS/RIS vendors bring voice recognition to radiology

Voice recognition has found an unlikely ally: PACS. Every major imaging and PACS vendor now offers, or is on the verge of launching, digital dictation capabilities integrated with their radiology workstations. Some even go a step further and voice-enable systems to respond to spoken commands.

The goal is to help radiologists streamline workflow and improve efficiencies, and the issue of whether the technology is accurate enough seems to have become moot.

A key player in this latest image management trend is Talk Technology. In 1996, when IBM introduced MedSpeak, the first commercially available continuous speech recognition product for radiology, Talk Technology began marketing the product. But MedSpeak had its drawbacks. Recognizing an opportunity to improve on the core technology, Talk Technology licensed IBM's speech recognition engine for radiology and built its own workflow application around it.

"The value for us is in the workflow," Mardini said. "We see this as a clinical reporting solution for radiologists to help make them more efficient."

Talk Technology claims installations of its Talk Station product at more than 200 hospitals, primarily for radiology. The company has formed alliances with more than a dozen PACS and RIS firms, including Agfa, Siemens, GE, Philips, Marconi, IDX, and Sunquest. It has modified Talk Station to be engine independent and has customized the original radiology package for use in pathology and emergency medicine.

"Our vision is that Talk Station will be the de facto standard speech and reporting interface inside of PACS/RIS workstations," Mardini said. "Our PACS partners are recognizing that voice recognition and clinical reporting aren't their focus, and they see that they can embed Talk Station to handle this."

In fact, PACS vendors say the transcription savings from Talk Station can reduce the overall return on investment from five to four years, according to Mardini. Turnaround times are reduced from 24 hours to one hour.

"I can walk into any site and save them 50% of their transcription costs," he said. "On average, 75% to 80% of reports at our customer sites are done without the aid of transcriptionists."

Numbers like these are what helped attract the attention of Talk Technology's PACS and RIS partners. Ongoing improvements in the technology and increasing pressure for radiologists to speed turnaround times have improved receptiveness among physicians as well.

"In the U.S. the focus is more on rapid turnaround of clinical decisions and getting the job done as quickly as possible," said Milan DiPiero, product manager for enterprise PACS at Philips Medical Systems. "Talk Technology has a lot of domain-specific knowledge that would take a lot of time for us to create or capture ourselves."

Philips offers Talk Station in conjunction with its own speech engine (the Speech Mike) in its PACS workstations for radiology, cardiology, and pathology. In addition, given Philips' core competencies in speech processing technologies and the strong installed base of Speech Mike among radiologists, the company is working to voice-enable its PACS workstations.

Other PACS vendors working with Talk Technology include GE Medical and Marconi Medical. GE showed a work-in-progress voice application at the last RSNA meeting that combined Talk Station with the PathSpeed workstation; the company's goal is to reduce the number of PCs sitting in the radiology viewing room and to streamline the ability to access and update images and reports.

GE is also in the early phases of integrating the Talk Technology package with Applicare's RadWorks, which is GE's lower-end PACS product. This would make the same integrated voice recognition capabilities available to smaller clinics and hospitals.

Marconi, the latest addition to Talk Technology's pool of PACS partners, is integrating Talk Station v.2 with its PACS workstations and Web-based results distribution capabilities to enable radiologists to dictate, edit, sign, and distribute coded reports in real-time or in batch mode. Marconi's results server automatically selects key images based on the report content of the dictation, eliminating the need for manual image marking.

"People will adapt to VR because of the time and cost savings associated with it and the improvements in patient care," said Vickie Sims, marketing manager for Marconi Medical. "Transcription is a very large part of the cost of the total study done on a patient. The scan lasts about 60 seconds, while the rest of the process takes two to four days."

Ultimately, Mardini sees voice recognition playing a central role throughout the hospital. Talk Technology is developing a handheld digital dictation product for enterprise applications that should be available by next year's HIMSS meeting.

"One thing is for sure beyond death and taxes: in the next five years, everything will use speech recognition for a large number of document creation and processing applications," Mardini said. "Three years from now, no one's going to be talking about who's got the best core engine. It will be the applications, and the enabling technologies that make it possible to build those applications."

Ms. Kincade is editor of Health Networking News, a biweekly business newsletter covering the medical information technology industry published by Diagnostic Imaging.