"Expert Systems and Open Systems in Medical Artificial
Intelligence".
John H. Frenster, M.D., FACP
Physicians' Educational Series,
Atherton, CA 94027-5446
CONTENTS:
Abstract:
Expert systems provide pre-selected rules for decision-making
within specialized domains of knowledge, but are limited by the fixed choices
and by the date of the expert opinions embodied in the decision rules.
Open systems, by contrast, allow user formulation both of the potential
decisions as well as of the decision elements leading to each decision,
while quantitating the contribution of each element to each potential decision.
Open systems can then calculate the best decision, the analysis of variance,
the logical inconsistency of selections, and the performance consistency
of ratings by utilizing the mathematical techniques of the analysis of
queueing and renewal with matrix cognition.
Medical Expert Systems:
Medical expert systems have evolved to provide physicians with
both structured questions and structured responses within medical domains
of specialized knowledge or experience (1). The structure is embodied in
the program on the advice of one or more medical experts, who also suggest
the optimal questions to consider, and provide the most accurate conclusions
to be drawn from the answers the physician chooses. In software programs,
these decision sequences are represented in clauses of the form: "If...,
Then..., Else...", with final else having positive value in the closed
system of the program (5). Although the physician is free to select any
one of the choices offered in each clause, the physician is limited to
the choices offered by the expert in writing the program. The program is
thus limited by the fixed input from the expert at the particular time
of formulation. If the physician has new questions or new data, a medical
expert system program will not be able to accomodate the physician. It
is for this basic reason that open system programs have been developed
to meet the new needs of the user (5), with the contrast being paraphrased
as: "Expert Systems are by experts; Open Systems are for experts".
Medical Open Systems:
Open systems in medical artificial intelligence allow physician
initiative in formulating both the decision outcomes and the decision elements
leading to such outcomes. Each physician user is allowed to formulate 2-10
potential outcomes and 1-10 contributing elements before proceeding with
the analysis (3). The physician then compares the decision elements for
relative importance in a pairwise response matrix that generates an eigenvalue
in a system of matrix cognition (6). The final decision outcome is then
calculated from the weighted decision elements, and the response eigenvalue
is used to calculate the response variance, the logical inconsistency of
the selections, and the performance consistency of the ratings (7) for
each user on each program run. The analysis of queueing and renewal within
human systems (8) has proven useful in generating both potential decision
outcomes and in identifying decision elements in awide variety of clinical
programs (3).
The Phases of Clinical Care:
The clinical care of a particular patient often proceeds in
distinct phases, such as diagnosis before therapy, or prevention of disease
before onset of disease, or rehabilitation of the patient after therapy
of the patient (2). The analysis of queueing and renewal within human systems
(8) has permited the identification of both decision elements and potential
decisions in at least 10 such distinct phases of clinical care (3):
| Phase of Care | Decision Elements | Potential Decisions |
| Prediction of Disease | Risk Factors Present | Predicted Disease |
| Prevention of Disease | Motivation of Patient | Preventive Measures |
| Diagnosis of Disease | Diagnostic Findings | Disease Diagnosis |
| Staging of Disease | Staging Factors Present | Disease Stage |
| Therapy of Patient | Pathologic States Present | Therapy Selected |
| Rehabilitation of Patient | Residual Defects Present | Schedule Selected |
| Health Status of the Patient | Specific Load Tolerances | Specific Capacities |
| Counseling of the Patient | Specific PatientConcerns | Specific Advice |
| Advocacy for the Patient | Specific Dangers to Patient | Specific Defenses |
| Financing for the Patient | Specific Medical Expenses | Specific Funding |
Following the identification of both decision elements and potential decisions in these distinct phases of clinical care, it became possible to develop open systems for decision-making, utilizing a system of matrix cognition (6) for each clinical phase (3), and allowing the teaching of the mathematical approach for physicians in each of these distinct clinical phases (2).
Matrix Cognition:
The development by Saaty of a mathematical analysis of pairwise
comparisons of user responses (6) has continued to exert a profound influence
on computer applications designed to enhance interactive human cognition
(7). The matrix methods developed by Saaty permit a reduction of many human
evaluations to a long series of pairwise comparisons, in which the accumulating
results are stored for matrix calculation while the user can focus serially
on distinguishing only two qualities or two quantities (3). Earlier studies
by Miller (9) had demonstrated the limited capacity of all humans to process
more than
7 plus or minus 2 separate foci of information at any one time,
and the development of matrix cognition with computer storage of user evaluations
for later matrix calculation of the response eigenvalues promises to maximize
the human capability for two-point discrimination by interaction with such
open systems (5).
Mathematical Logic:
John McCarthy has recently written on the importance of viewing
artificial intelligence "as a branch of computer science rather than as
branch of psychology" (10), but the two fields are rapidly converging,
and further benefitting from the introduction of mathematical techniques
and logical rigour, which now seem to permit human cognition over an extended
time frame and complexity net. Whether "the study of Artificial Intelligence
may lead to a mathematical metaepistemology analogous to metamathematics"
(10) is problematical, but the development of open systems and their use
by inquistive physicians may eventually help our patients most of all.
References:
1. Luger GF, Stubblefield WA, Artificial intelligence and the
design of expert systems, Redwood City, CA Benjamin/Cummings Publ. Co.
1989.
2. Frenster JH, Physicians' 1,2,3,4,5: Teaching physicians to think mathematically about each of their patient's problems. Innovations in Medical Education 1987; vol. 12, pages 87-88, Assoc. Am. Med. Colleges, Washington, DC.
3. Frenster JH, Expert systems and open systems within medical decision-making, Clin Research April, 1989 Vol 37. (Abstract).
4. Kant E, Interactive problem solving using task configuration and control. IEEE-Expert 1988; Winter: 36-49.
5. Hewitt C, Artificial intelligence: the challenge of open systems.
BYTE 1985, April: 223-273.
6. Saaty TL, A scaling method for priorities in hierarchical structures. J Math Psychology 1977; 15:234-281.
7. Saaty TL, Vargas LG. Inconsistency and rank preservation. J Math Psychology 1984; 28:205-214.
8. Frenster JH, Analysis of queueing and renewal within human systems. Nature 1965; 207:1139-1140.
9. Miller GA, The magical number seven, plus or minus two: some limits on our capacity for processing information. Psychol Rev 1956; 63:81-97.
10. McCarthy J, Mathematical logic in artificial intelligence. In Graubard SR,ed. The artificial intelligence debate: false starts, real foundations. MIT Press, Cambridge, MA 1988: 297-311.
1. Frenster JH, Matrix Cognition in Medical Decision-Making, Proc. Am. Assoc. Medical Systems and Informatics, vol. 7, pp. 131-134, (May, 1989).
Top of Page - Medical Systems Analysis - Euchromatin Network - Forums -Current Research -
matrixcognition: "Computer-Assisted Decision-Making".