Category : Various Text files
Archive   : NLM-INFO.ZIP
Filename : EXPERSYS.TXT

 
Output of file : EXPERSYS.TXT contained in archive : NLM-INFO.ZIP
NATIONAL INSTITUTES OF HEALTH
National Library of Medicine
NOVEMBER 1990
Expert Systems Program

Expert systems are computer programs that combine knowledge of a
particular subject area with inferencing mechanisms enabling them to
use this knowledge in problem-solving situations. An artificial
intelligence research program concentrating in expert systems was
established at the Lister Hill National Center for Biomedical
Communications (LHNCBC) in 1984. The objective of the Expert Systems
Program is to facilitate computer-assisted access to knowledge. The
fact that this knowledge may reside in different forms, in different
places, on different media, with different structures and naming
conventions, has engendered a particular focus on multimedia expert
systems.

The flagship project of the Expert Systems Program is the continuing
development and evaluation of the AI/RHEUM expert consultant system
by the Staff of the Center's Computer Science Branch. The AI/RHEUM
knowledge base has been updated and extended and linked to a new
rheumatology image library. A knowledge base compiler and a case
data editor have been designed, coded and tested. Useful critique by
advisory groups has resulted in the restructuring and streamlining of
the system's data entry process in preparation for testing in
clinical settings in Utah and Missouri. In its current state, the
AI/RHEUM diagnostic system contains, in its knowledge base,
information on 32 rheumatologic diseases. It reasons from 913
patient findings (basic information such as signs, symptoms,
laboratory tests and radiographic observations) through many hundreds
of intermediate hypotheses and criteria "decision tuples" (major
elements, minor elements, requirements and exclusions) to these 32
disease conclusions. It has 468 text definitions (the "Tell Me More"
knowledge source) available in a fraction of a second to explain
those patient findings which might not be familiar to its intended
users.

In addition, the system offers direct access to a "Show Me More"
videodisc image bank illustrating specific rheumatologic findings.
The first of the rheumatology image library videodiscs contained
about 1,900 slide images. The current video image bank contains
6,347 such still frames and 15 minutes of brief motion video
sequences to help physician users accurately make patient
observations difficult to illustrate statically. Scripting has just
been completed for voice-over narration for the video motion
sequences, which will be the final stage in the production of NLM's
rheumatology image library. Carrying further the ideal of access to
knowledge in other forms in other places, AI/RHEUM was the first
medical expert system to offer its user immediate searching of NLM's
MEDLINE" database for current information. The search is fully
automated, as AI/RHEUM passes stored queries to the "search engine"
of the Grateful Med" program. As of this writing, AI/RHEUM can make
136 automated MEDLINE searches on behalf of users needing current
information.

AI/RHEUM is the best known of a series of knowledge-based medical
consultant systems using the criteria table form of knowledge
representation pioneered by NLM researchers. The power, simplicity
and flexibility of this representation are augmented by a new expert
system shell written at NLM for the development of criteria-based
reasoning systems. The shell, called "CTX" for its use in criteria
table expert systems, will shortly go into beta-test phase at several
collaborating sites. CTX writes dBASE III- and Clipper-compatible
case data files. It is extensible and maintainable, and potentially
a useful building block for integration into complex projects which
need decision-support components. The new shell allows direct
coupling of video image libraries to expert systems and to programs
such as the Grateful Med Search Engine' for fully automated dialout
and searching of mainframe databanks. Several software tools written
as adjuncts to the CTX shell provide utilities assisting the
developer in handling multi-thousand-frame video image banks and in
automating the performance evaluation of CTX-based consultant systems
against benchmark sets of test cases.

CTX can analyze its reasoning on individual cases or on sequences of
cases, greatly simplifying the process of debugging changes to the
knowledge base and allowing the user to ask the equivalent of "Why
not X?", when a particular disease conclusion he or she thought was
likely was not triggered. It works with files created by standard
text editors, so the subject matter expert himself or herself can
work directly with the criteria table file to modify a knowledge
base. To help developers assess the utility and usage level of the
alternate knowledge sources, CTX can log to a transaction file, user
requests for help frames and for Tell-Me-More, Show-Me-More or
Search-for-More assistance with specific findings. The shell, with
its explicit multimedia links to knowledge sources in different forms
in different places, even on different machines, is one focus of the
overall Expert Systems Program goal of providing users with access to
knowledge. Its unique combination of capabilities can help
developers build consultant systems in any domain which lends itself
to the criteria form of knowledge representation.

The AI/RHEUM system has been carefully tested with several thousand
cases. But the evaluation of medical expert systems like AI/RHEUM is
a difficult problem for which no generally accepted paradigm has yet
been developed. Members of the Expert Systems Program are working
with other NLM staff and with nationally known evaluation specialists
to develop a general methodology for the evaluation of medical expert
systems. The AI/RHEUM diagnostic system is being used as the
specific vehicle for the testing of this methodology in the clinical
setting in an NIH-funded evaluation program of several years'
duration.

The AI/COAG hemostasis consultant system reported in prior years,
became in 1990 the basis for a collaboration between LHNCBC's Expert
Systems Program and the Medical Informatics Program and the
Department of Laboratory Medicine at Yale University. Building on
the AI/COAG knowledge base, the Yale groups have developed a
hemostasis advising system for in-house testing and potential
distribution.

The COACH' expert searcher system, the most recent of the Expert
Systems Program projects, is running now in prototype form. It works
with NLM's Grateful Med program to help improve retrieval from
MEDLINE for users who have gotten too little, too much, or poorly
focused search results. COACH will use Unified Medical Language
System knowledge sources such as the Metathesaurus', and the
inference engine from CTX to emulate some of the actions of an expert
human searcher diagnosing and responding to retrieval problems.

The interactive videodisc exhibit on "Artificial Intelligence in
Medicine," produced by Expert Systems Program staff with LHNCBC's
Audiovisual Program Development Branch, recently completed a three-
year national tour after being seen by nearly three million visitors
at eight major museums of science. This exhibit, one component of a
larger exhibition called "The Age of Intelligent Machines," opened at
the Museum of Science in Boston in 1987. It has spent three-month
periods there and at such other institutions as the Franklin
Institute in Philadelphia, the Fort Worth Museum of Science and
Industry in Texas, the California Museum of Science and Industry in
Los Angeles, and the Museum of Science and Industry in Chicago.

Recognizing the importance of fostering medical informatics research
by helping to create a "seed crop" of young researchers, Dr. Lawrence
Kingsland of the Expert Systems Program serves as coordinator for an
eight-week NIH "Medical Informatics" elective for third-year and
fourth-year medical students. Thirteen students from medical schools
across the U.S. completed the elective in 1990. The course included
a seminar series of 37 ninety-minute lectures, independent research
projects under the direction of NIH preceptors, and oral and written
presentations of research results. Several of these extremely
bright, highly motivated students have made important contributions
to Expert Systems Program projects.

For further information, contact:

Chief, Computer Science Branch
L ister Hill National Center
for Biomedical Communications
N ational Library of Medicine
8 600 Rockville Pike
B ethesda, Maryland 20894



  3 Responses to “Category : Various Text files
Archive   : NLM-INFO.ZIP
Filename : EXPERSYS.TXT

  1. Very nice! Thank you for this wonderful archive. I wonder why I found it only now. Long live the BBS file archives!

  2. This is so awesome! 😀 I’d be cool if you could download an entire archive of this at once, though.

  3. But one thing that puzzles me is the “mtswslnkmcjklsdlsbdmMICROSOFT” string. There is an article about it here. It is definitely worth a read: http://www.os2museum.com/wp/mtswslnk/