ASTM E2364-04

NOTICE: This standard has either been superseded and replaced by a new version or withdrawn.
Contact ASTM International (www.astm.org) for the latest information
Designation:E2364–04
Standard Guide to
Speech Recognition Technology Products in Health Care
This standard is issued under the fixed designation E2364; the number immediately following the designation indicates the year of
original adoption or, in the case of revision, the year of last revision. A number in parentheses indicates the year of last reapproval. A
superscript epsilon (´) indicates an editorial change since the last revision or reapproval.
1. Scope 3. Terminology
1.1 This guide identifies system types and describes various 3.1 Definitions:
features of speech recognition technology (SRT) products used 3.1.1 acoustic model, n—phoneme map of user.
to create the healthcare record. This will assist users (health 3.1.2 authentication, n—the process of confirming author-
information professionals, medical report originators, admin- ship of an entry or of a document, for example, by verifying
istrators, medical transcriptionists, speech recognition medical with a written signature, identifiable initials, computer key, or
transcription editors (SRMTEs), system integrators, support other methods.
personnel, trainers, and others) to make informed decisions 3.1.3 author, n—person responsible for content of text file.
relating to the design and utilization of SRT systems. 3.1.4 back-end system, n—delayedprocessingfordocument
1.2 This guide does not address the following items: completion.
1.2.1 System and data (voice and text) security. 3.1.5 compound file, n—a file containing recorded voice
1.2.2 Administrative processes such as authentication of the with its transcribed text.
document, productivity measurements, etc. 3.1.6 context, n—a long list of vocabulary words and
phrases used for the particular subject matter, with their
2. Referenced Documents
spellings and pronunciations, statistical information about
2.1 ASTM Standards: usageofeachwordaloneandincombination.Forexample,the
E1902 Specification for Management of the Confidentiality
context may include the number of times that “right,”
and Security of Dictation, Transcription, and Transcribed “Wright,” “turn right,” “right turn,” “right hand,” and “Mr.
Health Records
Wright” occur in a body of text. It also includes grammar and
E1985 Guide for User Authentication and Authorization styleinformation.Languagemodel,lexicon,topic,andvocabu-
E2084 Specification forAuthentication of Healthcare Infor-
lary are terms that are all used synonymously with context.
mation Using Digital Signatures
3.1.7 digital signature, n—data associated with, or a cryp-
E2184 Specification for Healthcare Document Formats tographic transformation of, a data unit that allows a recipient
E2185 Specification for Transferring Digital Voice Data
to prove the source and integrity of the data unit and protect
BetweenIndependentDigitalDictationSystemsandWork- against forgery, for example, by the recipient.
stations
3.1.8 edit, v—to review the document while listening to the
E2344 Guide for Data Capture through the Dictation Pro- originator’s recorded voice and reading the associated tran-
cess
scribed text (compound file), checking for recognition errors
2.2 Other Documents: and correcting document formatting and other inconsistencies.
Resource Interchange File Format (RIFF) Standard
When the SRMTE is not the originator, the SRMTE may need
to flag the document for originator/author clarification of
unclear content or intent.
This guide is under the jurisdiction of ASTM Committee E31 on Healthcare
3.1.9 encryption, n—the process of transforming plain text
Informatics and is the direct responsibility of Subcommittee E31.15 on Health
Information Capture and Documentation. (readable) into cipher text (unreadable) for the purpose of
Current edition approved June 1, 2004. Published July 2004. DOI: 10.1520/
security and privacy.
E2364-04.
3.1.10 front-end system, n—a system incorporating real-
For referenced ASTM standards, visit the ASTM website, www.astm.org, or
time recognition and may include real-time self-editing by the
contact ASTM Customer Service at service@astm.org. For Annual Book of ASTM
Standards volume information, refer to the standard’s Document Summary page on
originator.
the ASTM website.
3.1.11 language model, n—context specific to medical spe-
Withdrawn. The last approved version of this historical standard is referenced
cialty, user, or practice setting.
on www.astm.org.
Copyright © ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959, United States.
E2364–04
3.1.12 lossless compression, n—a lossless compression re- 3.2.3 RIFF—resource interchange file format
duces the amount of data required to represent the original 3.2.4 SRT—speech recognition technology
voice file but has no impact on sound quality. The original file
can be replicated precisely at any time. 4. Significance and Use
3.1.13 lossy compression, n—a lossy compression loses
4.1 This guide is intended to provide general guidelines
someinformation,resultingindegradationofthesoundquality
toward the design and utilization of SRT products used for
inherent in the original voice file and an inability to precisely
healthcare documentation. It is intended to recommend the
regenerate that original file.
essential elements required of SRT systems in healthcare.
3.1.14 microphone, n—an instrument whereby sound waves
4.2 This guide will not identify specific products or make
are caused to generate or modulate an electric current usually
recommendations regarding specific vendors or their products
forthepurposeoftransmittingorrecordingsound(asspeechor
or services.
music).
4.3 A well-edited SRT document may result in improved
3.1.15 microphone element, n—the component within the
quality over current methods of documentation, that is, hand-
microphone that does the actual conversion from sound waves
written notes and improved productivity over traditional dic-
to electrical signals.
tation and transcription.
3.1.16 natural language processing, n—method used in
4.3.1 Faster turnaround times.
artificial intelligence to process and derive interpretation of
4.3.2 Legible documentation over handwriting has many
human language.
advantages:
3.1.17 networked system, n—system connected to a net-
4.3.2.1 Improved patient care communication.
work.
4.3.2.2 Enhanced patient safety.
3.1.18 “normal” dictation, n—routine phrases or para-
4.3.2.3 Reduced malpractice risks.
graphs.
4.3.2.4 Facilitation of appropriate reimbursement.
3.1.19 originator, n—person who provides oral input or
4.3.3 For the medical transcriptionist and/or SRMTE, de-
dictation, not necessarily the person responsible for the con-
creased repetitive stress injuries, such as neck, arm, wrist, and
tent.
heel pain.
3.1.20 phoneme, n—smallest unit of sound in a spoken
4.3.4 Facilitation of cost controls related to document
language.
completion.
3.1.21 prompts, n—remindersprovidedinordertocomplete
4.3.5 Better utilization of medical language skills of MTs as
a task.
productivity is not limited by keyboarding skills.
3.1.22 real-time recognition, n—simultaneous speech-to-
text transcription.
5. Speech Recognition Technology Systems
3.1.23 RecO—speech recognition error
5.1 Speech recognition technology (SRT) is designed to
3.1.24 RIFF file, n—Resource Interchange File Format
capture voice and transcribe that speech into text. This can be
(RIFF) is self-descriptive; that is, the voice file format is
done by a single user working at a standalone computer or by
defined within the file.
a large group of users working on a network. Another method
3.1.25 speech recognition, n—computerized transcription
is processing a pre-recorded digital voice file through an SRT
of speech to text.
system, with the resulting text and/or SRT engine being edited
3.1.26 speech recognition medical transcription editor,
by the MTE.
n—medical transcriptionist who edits compound files and/or
5.2 Speech recognition technology system workflow.
the SRT language model.
5.2.1 Front-end speech recognition process involves:
3.1.27 SRT engine, n—speech recognition processor.
5.2.1.1 Recording the voice.
3.1.28 standalone system, n—system not connected to a
5.2.1.2 SRT transcription of the voice file to text.
network.
5.2.1.3 Editing may be done by the originator and/or
3.1.29 synchronization, v—having voice and text matched
SRMTE.
such as in a point-and-play manner.
5.2.1.4 Compound file may be saved as an option.
3.1.30 text file, n—a file that contains text message.
5.2.1.5 Text file can be printed, archived, transmitted, or
3.1.31 voice enrollment, n—the process whereby a user
integrated into an electronic health record.
reads aloud selected text so the SRT software can map or
5.2.1.6 Update the SRT context for RecOs and new termi-
record the user’s speech sound pattern (phonemes).
nology.
3.1.32 voice file, n—digitalized audio message representing
5.2.2 Back-end speech recognition process involves:
voice input.
5.2.2.1 Recording the voice.
3.1.33 voice macros, n—stored keystrokes that are activated
5.2.2.2 Transmitting the voice file to the speech recognition
by a voice command.
engine.
3.1.34 WAV, n—voice file format.
5.2.2.3 SRT transcription of the voice file to text.
3.2 Acronyms: 5.2.2.4 Saving the voice and text as a compound file.
3.2.1 MT—medical transcriptionist 5.2.2.5 Routing the compound file to the SRMTE.
3.2.2 SRMTE—speech recognition medical transcription 5.2.2.6 Editing done by the SRMTE.
editor 5.2.2.7 Saving the text file.
E2364–04
5.2.2.8 Returning the edited text file to the originator for 5.2.4.6 The following scenarios are offered to give the
authentication. reader examples of how these systems work. They are not
intendedtorepresenteverypossiblescenarioforthesesystems.
5.2.2.9 SRMTE updates the SRT context for RecOs and
new terminology. (1) Six radiologists simultaneously dictate at individual
workstations. Each voice file is routed to a recognition server,
5.2.3 Standalone SRT System:
or the processing may take place on each workstation, with
5.2.3.1 Only one person at a time can use a standalone
information regarding the originator’s specialty and identifica-
system.
tion, allowing the recognition server to load the corresponding
5.2.3.2 Context is limited by the hard drive space.
acoustic model and context. The voice file is processed by the
5.2.3.3 Editing is done locally, at the point of input, either
SRT engine and the resulting compound file (voice and text
by the originator or by the SRMTE.
files) is routed to the SRMTE for editing. The SRMTE may
5.2.3.4 Input devices.
also be responsible for editing the SRT context.
(1) Noise-canceling SRT microphones.
(2) A hospital has 300 healthcare providers dictating into
(2) Handheld digital recorders.
portable handheld digital recording devices from the hospital
(3) Digital dictation systems.
and several remote satellite clinics, or dictation may take place
(4) Telephones.
on individual workstations. The voice files are encrypted and
5.2.3.5 The following scenarios are offered to give the
securely transmitted to the digital dictation system of a
reader examples of how these systems work. They are not
contracted transcription company. Each voice file is routed to a
intendedtorepresenteverypossiblescenarioforthesesystems.
recognition server, or the processing may take place on
(1) A radiologist (originator) dictates into a microphone
workstations, with information regarding the originator’s spe-
connected to a personal computer running an SRT program.
cialtyandidentification,allowingtherecognitionservertoload
The voice is translated to text in real time. The originator edits
the corresponding acoustic model and context.The voice file is
the text and/or the SRT context.
processed by the SRT engine and the resulting compound file
(2) A family practitioner dictates into a personal computer
(voice and text files) is routed to the SRMTE for editing.
throughout the day. Each compound file is saved and then,
SRMTEs working both in the office and remotely receive
using the same computer, the SRMTE edits the text, listening
recognized compound files via encrypted Internet transmis-
to the recorded voice as necessary for clarification. The
sions. The editing is performed on standalone computers and
SRMTE may also be responsible for editing the SRT context.
the encrypted text files are returned. The SRMTE may also be
(3) A group of cardiologists dictate into handheld digital
responsible for editing the SRT context.
recording devices throughout the day. The voice files are
transmittedfromtherecorderstoacomputerandrecognizedby
6. Training
the SRT engine, using the cardiology context and each physi-
6.1 Originators:
cian’s acoustic model. Once recognized, each text file is edited
6.1.1 Voice enrollment and proper position of microphone
by the SRMTE. The SRMTE may also be responsible for
and proper placement of microphone element.
editing the SRT context.
6.1.2 Build customized language model.
5.2.4 Networked SRT System:
6.1.3 Build “normal” dictations per user.
5.2.4.1 On a networked system, all files containing recorded
6.1.4 Develop skill sets.
dictation(voicefiles)aretransmittedtoaserver,wherethefiles
6.1.4.1 Proper correction technique for a RecO.
are queued up for recognition. The compound files are then
6.1.4.2 Navigation/mobility skills for moving around in the
routed to the SRMTE for editing.
document.
5.2.4.2 A networked system is designed to allow multiple
6.1.4.3 Editing skills specific to SRT products.
originators and SRMTEs to work simultaneously. The voice
6.1.4.4 Editing language model.
files are recognized on a server or at the workstation(s) and the
6.2 Speech Recognition Medical Transcription Editor:
resulting compound files are routed to the SRMTE for editing.
6.2.1 Voice enrollment and proper position of microphone
5.2.4.3 Contexts.
and proper placement of microphone element.
(1) The networked system may be programmed for a single
6.2.2 Build customized language model.
medicalspecialtyorsubspecialty,suchasradiology,pathology,
6.2.3 Build “normal” dictations per user.
family practice, physical therapy, or emergency medicine.
6.2.4 Develop skill sets.
(2) A networked system may also be programmed with
6.2.4.1 Proper correction technique for a RecO.
many contexts or language models so originators from many
6.2.4.2 Navigation/mobility skills for moving around in the
different medical specialties can use it to improve speech
document.
recognition accuracy.
6.2.4.3 Editing skills specific to SRT products.
5.2.4.4 Editing may be done in the same facility, or the
6.2.4.4 Editing language model.
compound files may be sent to a remote SRMTE.
6.2.4.5 Start and stop audio file.
5.2.4.5 Input devices.
6.2.4.6 Identify a RecO.
(1) Noise-canceling SRT microphones.
7. Realities of Speech Recognition Tec
...