USE SPEECHPROFESSIONAL™ WITH OFF-THE-SHELF USB FOOT CONTROL

No need for a high-cost "certified" foot control" with this system.  Why pay more . . . .

SpeechProfessional™ software suite includes workflow manager, server-based speech processing, and dictation and transcription desktop utilities.  The included SpeechMax™ session file editor for dictation, transcription, and speech recognition supports dictation audio playback while highlighting text to verify accuracy.    It can be used with an off-the-shelf USB Infinity transcriptionist footpedal or other HID compliant USB device.  The company also offers a "2 for 1" package that includes two session file editor licenses plus USB foot control for only $289.  Call Deb at 219-662-3800 by January 11, 2008 and get this two-for-one package for just $189.  Offer valid for single purchase only.  Not available to employees, resellers, or with other discount packages.

PURCHASE DRAGON MEDICAL AND OTHER EDITIONS AND SAVE

Dragon NaturallySpeaking 9

Best buys for health care, law, business, schools, and government

Medical $1,199 (reg. $1,599)
Legal $899 (reg. $1,199)
Professional $699 (reg. $899)
Preferred $179 (reg. $199)
Medical Small Practice Edition $899 (reg. $1,199)
Medical Small Practice Edition 5 user Value Pack $3,899 (reg. $5,499)
Runtimes Medical $699 Legal $699 Professional $499*

* Dragon speech recognition dictation-only runtimes may be purchased from Custom Speech USA only for use with the company's products.  Electronic Help manual is included with the runtime, but not voice commands, printed manual, or headset microphone.  See other products for pricing and more information.

We also offer upgrades, value packs, bundles, and open license programs.  Call 219-662-3800 TODAY for a FREE quote, or use the Online Information Form to get more information about Dragon and other third-party solutions.

Say I Can teaching CDs for Dragon NaturallySpeaking also available.

ECONOMICS OF SPEECH RECOGNITION EDITING FOR RADIOLOGY

Radiologists are among the most common users of speech recognition.  However, William R. Reinus, MD, MBA, Temple University Medical Center (Philadelphia, PA) warned that radiologist productivity may be reduced by up to 25% if hospitals force radiologists to correct their own speech recognition reports.  Dr. Reinus also referred to potential lengthy review time by transcriptionist to listen to the dictation and locate words misrecognized by the speech engine.  In dealing with alternatives, he further cautioned that there may be limitations with use of structured dictation and templates, such as "modification of templates for studies with near-normal results."  Many clinicians, he said, "expect customized reports for their patients' studies, particularly more sophisticated studies." Economics of Radiology Report Editing Using Voice Recognition Technology (Journal of American College of Radiology December 2007)

Howard P. Forman, MD, MBA, a radiologist from Yale University Medical Center responded in a companion article that his institution's "transcription services, whether homegrown or outsourced, would take anywhere from 2 hours to 2 weeks to deliver a transcribed report. Even in efficient settings, reports might not be edited and signed on the same day as they were dictated." Re: "Economics of Radiology Report Editing Using Voice Recognition Technology" (Journal of American College of Radiology December 2007) He further indicated that he would prefer to edit his own reports, and send them off immediately to a referring clinician without waiting for correction by an
assistant.

SpeechProfessional™ software suite add-on for Dragon, IBM, and Microsoft speech recognition can make a radiology practice more efficient by dealing with the issues raised by the Dr Reinus and Dr. Forman.

1. Server-based speech recognition with editor correction.  With server-based systems and offline correction, there is no significant change in radiologist workflow.  The dictation audio is routed to a server-based system.  A preliminary report is generated by "computerized transcriptionist."   Speech recognition systems are up to 99% accurate.   Reports are edited by a transcriptionist editor and returned to the radiologist for approval and digital signature, as with manual transcription.  Use of foot control and WordCheck™, advanced text comparison techniques, described below, improve editor efficiency and reduce turnaround time.  The company's unique "dual-engine" and "multi-engine" approach has been reviewed in health care, legal, and general technology journals.

2. Real-time speech recognition available also.  In some settings, such as requests for a STAT reading, that rapid turnaround is desirable.  With our system a radiologist can use real- time or server-based technology, depending upon the situation and personal preference. 

3. Easy implementation of customizable, study-specific and doctor-specific structured dictation and templates. Structured dictation may be used with "fill-in-the-blank" electronic forms. Template reports may include preset reports that apply to common clinical situations or imaging findings, such as a "normal CT Brain".  With both structured dictation and templates speaker generation time, speaker or transcriptionist editing, and basic errors such as dropped words (such as "no") can decrease.  SpeechProfessional™supports both structured dictation and templates using both server-based and real-time speech recognition, as well as for manual transcription. These normals and templates may be customized for particular imaging studies or referring clinicians.

4. RIS/HIS and EMR integration. From the outset, the Custom Speech  workflow system with SpeechProfessional™ was designed to support easy integration with third-party solutions. Most recently, the company has integrated Dragon server-based speech recognition with a JAVA-based medical query system at Columbia University Medical Center in New York City. Similarly, a large multispecialty medical clinic in Billings, Montana has added Dragon server-based speech recognition as a back-end add-on to a DVI dictation and transcription workflow. A
gastroenterology test site in Queensland, Australia integrated back-end Dragon server-based
workflow to a third-party dictation system. Other projects have included speech recognition for a Larchmont, NY radiology imaging center, creation of custom templates for use with real-time speech recognition and preexisting database structure at the Los Angeles County Coroner's Office, and HL-7 integration with speech recognition system for a large Hagerstown, MD community hospital.  In each of these cases, Custom Speech expertise and experience, scripting tools, software development kit, and open database engine architecture facilitated rapid and successful integration with a third-party system. For more information see Case Studies.

5. Unique features to improve efficiency and privacy. Custom Speech provides state of the art technology AND novel features not available with other systems.

****  WordCheck™ using dual or multiple speech engines for faster review of dictation audio. U.S. and foreign patented SpeechMax™ WordCheck feature can reduce review time by identifying probable errors and highlighting these for review by the transcriptionist. It does this by comparing synchronized text output by two or more speech recognition engines. Differences indicate possible errors. Agreement suggests accuracy. If there is 100% speech engine accuracy, there are no differences between the speech engines. With no differences, audio review time of a 1 minute or 1 hour file could be reduced to essentially zero before sending the text for physician approval. Where accuracy is less than 100%, differences will require review. But by tabbing through the document listening primarily to differences where the speech recognition engines are relatively accurate, the correctionist can shorten audio review time and more quickly produce an accurate document for review.  

****  Speaker-specific speech recognition for high accuracy.   Conventional speaker-dependent speech recognition systems use speech from many different speakers with different accents to create a speaker-independent speech user profile. Enrollment training (reading a 15 minute or half hour script) adapts the speaker-independent speech user profile to make it more speaker dependent.  The SweetSpeech™ speaker-specific model is based upon the speech, accent, background and channel noise, and vocabulary of only a single speaker. It reflects actual accent, speaking habits, and actual noise of the system, not a mathematical approximation of it based upon adaptive training. As a result, the system may be highly accurate. "Initial results show that by changing the focus to MSSR, word error rates can drop very significantly. In comparison with speaker-adaptive speech recognition system, massively speaker-specific speech recognition also performs better since model parameters can be tuned to be suitable to one particular individual." Y. Shi & E. Chang, "Studies in Massively Speaker-Specific Speech Recognition" (IEEE 2004)

****ScrambledSpeech™ to protect patient confidentiality. Department of Veterans Affairs Office of Inspector General, Audit of Veteran's Health Administration's Acquisition of Medical Transcription Services (VA Office of Inspector General June 14, 2006) audit showed that confidential VHA information was being transcribed overseas and that terrorists and others potentially had access to medical data identifying U.S. service members and their families.  Using with SpeechMax™ session file editor, to protect privacy, a user selects a speech recognition or manually-transcribed session file or dictation audio. The user divides the parent session file into shorter two or more child session files. Order of audio or text of each child session file is scrambled. The operator sends each child session file to a different transcriptionist or editor for processing. The operator then merges and unscrambles completed child session files to generate final document or report. The major benefit is that it limits transcriptionist or editor access to confidential information.
This can benefit any HMO, hospital, or health system that outsources transcription.  View the video demo. And the session file editor software supports lock/unlock of session files to prevent unauthorized editing of the final document (document protection).

HOW DOES TEXT TO SPEECH WORK?

Best-selling acWAVE™ audio conversion software includes FREE text to speech.  Use with AT&T Natural Voices Text to Speech and other Microsoft SAPI 5.x compatible voice fonts.  Using this tool, you can generate voice messages, telephone prompts, and other life-like sounding records.  A June 2007 SmartComputing article explains how the technology works.

"Speech synthesis, or the act of generating artificial speech using machines or software, is the fundamental technology behind TTS.  Research on the subject predates modern computing.  Homer Dudley’s speech synthesis machine, named Voder, could generate human-like sounds, and with the help of a trained operator, serenaded attendees at the 1939 World’s Fair with a warbled rendition of “Auld Lang Syne.” I n the 1950s, scientists used electronics designed to emit sounds at frequencies similar to those produced by the human vocal tract, but the results still sounded thoroughly inhuman.

Several decades later, Bell Labs researchers took another approach, called concatenative synthesis.  This method employs a limited amount of prerecorded speech split up into its fundamental vowel and consonant sound units, or phonemes, which are then mixed and matched to produce unique words and sentences.  The result was a much-improved ability to produce intelligible speech on the fly, but because the number of phonemes being used was rather limited, these systems sounded less than convincing.

As computing hardware improves, TTS algorithms are becoming better able to search for and splice together a greater number of phonemes, which has led researchers to focus more attention on sounds as they occur in context with adjacent sounds, varying pitch in a given sentence, and the stressing of sounds in certain words.  As a result, current TTS software can mimic human speech in a much more believable way than its predecessors."

Click here to review the complete article.

Custom Speech USA, Inc.
3 North Court Street, Suite B365Crown Point, IN 46307
219-662-3800 (v)
219-662-3877 (f)
www.customspeechusa.com
info@customspeechusa.com
We teach computers how to listen®

Online Information Request Form
_____________________________________________________

You are receiving this copy of the Custom Speech USA, Inc. newsletter because you have expressed interest in our products and services.  If you have been sent a newsletter in error, or would like to be removed from the list, please click here. 

Price, terms, specifications, and availability are subject to change without notice.  Custom Speech USA, Inc. trademarks are indicated. Other marks are the property of their respective owners.  Click here for Online Privacy Policy. 

Dragon and NaturallySpeaking® are licensed trademarks of Nuance® (Nuance Communications, Inc.)