Talking about a revolution

Neil Kelly

Speech recognition technology has been available commercially for around ten years, attracting many early adopters at a young stage of development when results were distinctly unimpressive. The scene is changing rapidly however.

Simon Howard, Dragon Professional and Healthcare Solutions EMEA director for Nuance Communications, believes that speech recognition was a victim of its space age nature. The technology did work in the early days, but the necessary processing power and RAM were not available.

Today, entry-level PCs are more than sufficient for the task. Software with high-end algorithms to cancel out noise combined with wireless noise-cancelling headsets enable a high level of accuracy.

Current market

Strategically, Nuance has targeted the healthcare market and offers products affordable to individual consultants or GPs at one end of the scale, all the way up to trust-wide/enterprise packages at the other. The first upgrade is a simple move from an analogue process, where a human listens to a tape and types out text, to a digital process, where a secretary receiving a speech-recognised piece of text becomes more an editor than a typist.

Howard acknowledges concerns recently raised by Unison that the technology will lead to job losses, but claims the real savings for trusts are to be made in faster and more accurate report turnaround times. He believes speech recognition will only require a different skill set.

In the primary care sector, Nuance markets its products via a number of partner companies. Vision Voice Dictate, developed by West Sussex-based Hands Free Computing and marketed by INPS (formerly In Practice Systems), essentially consists of the scripting mode in Dragon Speaking Professional interfacing with the INPS Vision GP system.

Vision Voice Dictate not only facilitates automatic dictation within Vision, but allows the user to control the whole interface. Hands Free Computing are looking to extend their offering to the web-based, centrally hosted version of Vision (Alt-GP).

In secondary care, says Howard, the recent incorporation of Dictaphone becomes significant. Dragon NaturallySpeaking is sufficiently robust for a 20-30 person department, but beyond that workflow-based Dictaphone offerings would be deployed. For example, Dictaphone's Enterprise Workstation (EWS) can be deployed up to regional level.

Essentially, nothing should have changed for consultants. They still dictate into a microphone in the same way, but now their text is sent to a recognition server which will 'speech recognise' the text, then route it to a human for any corrections. It is then sent back for the author to check and sign off.

Little 'training' of the software is involved and the author only need read a small amount of text to form a base profile. When the edited text comes back from the secretary the two versions are married up so that the voice engine is constantly learning the idiosyncracies of the author's voice.

Version 9 of Dragon Naturally Speaking, launched this month will, according to Howard, require no training at all. Paul Ricci, chairman and CEO of Nuance said: "With the new release, we have overcome the final obstacles to adoption and made speech as seamless as the keyboard and mouse."

Nuance will also be releasing a dedicated medical version of Dragon Naturally Speaking at the end of July, with 14 specialist medical vocabularies and will also soon be supporting Bluetooth.

Nuance claims a Dragon presence in over 200 NHS sites, including a 40% penetration of pathology labs and 25% penetration of radiology departments.

The clinician's point of view

Consultant radiologist Dr Richard Harries, has been using speech recognition technology for around five years and describes himself as an enthusiast. His interest was sparked by the shortage of radiology secretaries in his Grimsby, Scunthorpe and Goole NHS Trust, with reports sometimes taking weeks to be typed. The average turnaround time is now less than one week.

Dr Harries piloted the Phillips SpeechMagic product, then at version 4, before extending the technology to other departments. Initially, he was just converting his spoken words to text on a page, but he is now starting to develop some command controls to make better use of desktop integration between the trust's new Agfa picture archiving and communication system and its iSoft RIS. The RIS uses a digital dictation system with real-time speech recognition supplied by Soliton IT.

Dr Harries maintains that this systems integration has made a major contribution to patient safety as the manual method of taking data from separate systems made it all too easy to mix up patient notes.

Dr Colin Campbell, a senior radiologist in Glasgow Southern General Hospital, has been trialling Philips SpeechMagic for over a year. He tried other products before settling for SpeechMagic, as he found it easier to get started with and more accurate.

As the software is being used on a trial basis, it is not currently fully integrated with the department's radiology information system (RIS). This means Dr Campbell has to dictate his report, then check the text, then transfer it across to the RIS, then check it again and verify it. Even so, he says this still gets the reports out much faster.

He points to some 'niggling' problems, for instance the software represents millimetres in the American form as 'ml' and it is also difficult to get a forward slash recognised. Some interesting problems arise from his Scottish accent; as Speech Magic is designed to screen out hesitation noises, it can interpret a word like 'arm' as 'ah-um' and not type it. He finds the software struggles with the same small words that can be unclear to human typists, but is very good with the long technical words that it has in its dictionary.

No date has yet been set for integration with the RIS, and one obstacle is the current version of SpeechMagic not being compatible with the next upgrade of the RIS software. Dr. Campbell's hope is that the integrated product will be rolled out throughout the city within the next 12 months, with every radiology department in Glasgow using it routinely.