Brought to you by EarthWeb
IT Library Logo

Click Here!
Click Here!

Search the site:
 
EXPERT SEARCH -----
Programming Languages
Databases
Security
Web Services
Network Services
Middleware
Components
Operating Systems
User Interfaces
Groupware & Collaboration
Content Management
Productivity Applications
Hardware
Fun & Games

EarthWeb Direct EarthWeb Direct Fatbrain Auctions Support Source Answers

EarthWeb sites
Crossnodes
Datamation
Developer.com
DICE
EarthWeb.com
EarthWeb Direct
ERP Hub
Gamelan
GoCertify.com
HTMLGoodies
Intranet Journal
IT Knowledge
IT Library
JavaGoodies
JARS
JavaScripts.com
open source IT
RoadCoders
Y2K Info

Previous Table of Contents Next


3.4. CASE-BASED REASONING

While the neural network approach offers a useful tool for knowledge acquisition and representation, and is unlike human experts, this approach solves problems by reasoning from principles; that is, it explains the reasoning by reporting the string of deductions that lead from the input data to the conclusion. Medical doctors, however, solve new problems by analogy with old cases and explain reasons in terms of prior experience. Computer systems that solve by analogy with old ones are called case-based reasoning (CBR) systems (Rich et al., 1991). CBR systems solve problems by searching a collection of stored cases to find and retrieve the cases that most closely resemble (match) a newly presented case using match score. A CBR system draws its power from a large case library, rather than from a set of principles.

As seen in the Appendix, CBR has been applied to the following medical domains: analysis of skull structure, general surgery, radiation therapy, biochemistry, post-operative care, hemodynamics, nursing care, psychotherapy, heart disease, diagnostic imaging procedure, protein analysis, eating disorders, allergic rhinitis, jaundice, ICU care, HIV, radiology, kidney function, asthma, and leukemia. Of 233 expert systems listed in the Appendix, 30 systems (12.9%) were developed by CBR. The survey of expert systems developed in Germany (Kraut and Mann, 1996) also showed a similar percentage for CBR.

CBR enables the expert system to tap into a medical knowledge source (e.g., medical records and clinical case database) that is often more readily available in hospitals than outside practices. The neural network and CBR have complementary strengths. The neural network systems capture broad trends in the domain, while the CBR systems are good at filling in small pockets of exceptions.

3.5. OBJECT-ORIENTED PROGRAMMING

Object-oriented programming refers to all of their data structures as objects. Each object contains two basic types of information: information that describes the object and information that specifies what the object can do. In the expert system, each object has a declarative and procedural knowledge (Durkin, 1994). It provides a natural way of representing real-world objects. According to the survey conducted by Kraut and Mann (1996) of Germany, object-oriented programming was the second most frequently used method (54%) in medicine, next to rule-based reasoning (69%).

4. THE STATUS OF APPLICATIONS RESEARCH

This section reviews some of the well-known medical expert systems developed over the last 2 decades.

4.1. INTERNIST/CADUCEUS: AN EXPERT SYSTEM IN INTERNAL MEDICINE

The INTERNIST project was started in the early 1970s, and continues today under the name CADUCEUS (People, 1984). The goal of INTERNIST is to perform a diagnosis of the majority of diseases associated with the field of internal medicine. This, in itself, is an ambitious endeavor as there are hundreds of such diseases. However, not only is INTERNIST/CADUCEUS intended to diagnose each disease, it is designed to consider all the possible combinations of diseases that might be present in the patient. It is estimated that the number of such combinations is of the order of 10 to the 40th power. Consequently, as in the case of DENDRAL, we are faced with a problem that exhibits combinatorial explosiveness, a problem for which the heuristic approach is most appropriate.

4.2. MYCIN: AN EXPERT SYSTEM IN BLOOD INFECTIONS

MYCIN is, at this time, probably the most widely known of all expert systems thus far developed (Shortliffe, 1976), despite the fact that it has never been put into actual practice. Many early texts on expert systems have focused primarily on the MYCIN system and the project has served to substantially influence much of the subsequent work in the construction and implementation of expert systems. MYCIN was designed solely as a research effort to demonstrate how expert systems might actually be constructed for reasonably large and complex real-world problems. MYCIN is thus somewhat akin to INTERNIST/CADUCEUS in its purpose, except that it focuses on a far smaller number of diseases and thus requires a considerably smaller knowledge base.

The knowledge base of MYCIN contains the heuristic rules used by physicians in the identification of certain infections. EMYCIN (empty MYCIN)is the name given to MYCIN when this specific knowledge base is removed. In many cases, one may collect a knowledge base associated with a different domain and insert this into EMYCIN, where the result is a new, working expert system shell. The result of incorporating a knowledge base associated with pulmonary disorders into EMYCIN resulted in a new expert system known as PUFF.

4.3. PUFF: AN EXPERT SYSTEM IN PULMONARY DISORDERS

PUFF was developed in 1979, using the EMYCIN shell (Aikens et al., 1984). The purpose of PUFF is to interpret measurements related to respiratory tests and to identify pulmonary disorders. PUFF interfaces directly with the pulmonary test instruments used in such measurements. At the conclusion of the testing, PUFF presents the physician with its interpretation of the measurements, a diagnosis of the illness, and a proposed treatment scheme. The first version of PUFF had 64 production rules; a more recent version (coded in BASIC) has about 400 rules.

The validation procedure for PUFF was conducted by comparing its diagnosis with that of two expert pulmonary physiologists. The conclusions of PUFF and that of the physiologists were consistent in more than 90% of test cases. As a result, PUFF is now used on a routine basis.

One particularly interesting feature of PUFF is that the physical appearance is indistinguishable from conventional laboratory tools. The acceptance of PUFF by the medical profession may be based, in large part, on this perception; that is, PUFF is viewed as an ordinary piece of laboratory equipment, rather than as an intelligent (and possibly superior) competitor.

4.4. QMR: A MEDICAL DIAGNOSTIC EXPERT SYSTEM

Using the massive knowledge base first developed for INTERNIST, QMR (Quick Medical Reference) assists physicians in the diagnosis of an illness based upon the patient's symptoms, examination findings, and laboratory tests (Kane and Rucker, 1988). QMR, utilized at the University of Pittsburgh, incorporates over 4000 possible manifestations of diseases and is said to perform at a level comparable to practicing physicians.


Previous Table of Contents Next

footer nav
Use of this site is subject certain Terms & Conditions.
Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details.