[ CogSci Summaries home | UP
| email ]
http://www.jimdavies.org/summaries/
Bichindaritz, I. & Sullivan, K. M. (1998) Reasoning from Knowledge
Supported by
More or Less Evidence in a Computerized Decision Support System for
Bone-Marrow
Post-Transplant Care. AAAI Spring Symposium in Multimodal Reasoning.
Technical Report SS-98-04. 85--90
@Article{ BichindaritzSullivan1998, author =
{Bichindaritz, Isabelle and Sullivan, Keith
M.}, title =
{Reasoning from Knowledge Supported by More or Less Evidence in
a Computerized Decision Support System for Bone Marrow Post-Transplant
Care}, journal =
{American Association
for Artificial Intelligence Spring Symposium}, year = {1998}, volume =
{ }, pages =
{85--90},}
Author of the summary: Amy S. Hwang, 2006,
2ash1@qlnk.queensu.ca
Cite this paper for:
- System description:
A
knowledge-based computerized decision-support system applied to
the
long-term follow-up of bone marrow transplant patients at the
Fred
Hutchinson Cancer
Research Center.
[p85]
- Priority rules:
system can
prioritize the use of widely-recognized knowledge, over knowledge
with
limited recognition. [p85]
- Evidence-based
reasoning in
the context of bone-marrow transplants (BMT). [p86]
- Quality of practice:
based on
two main dimensions of knowledge. [p86]
- Case-based
reasoning, the
principle methodology of the system. [p86]
- Definition of a
problem to be
solved by the system, ct. [p86]
- Rule-based
reasoning, based
on the RETE algorithm of production systems. [p87]
- Reasoning RBR in
system
rule-based reasoning. [p87]
- Conflict set in
rule-based
reasoning. [p87]
- Working memory W,
within
rule-based reasoning. [p87]
- Knowledge
representation for
various types of knowledge within the system. [p88]
- System
knowledge-base. [p88]
- Two main categories
of
guidelines: diagnosis guidelines and treatment guidelines. [p88]
- The system
reasoning. [p89]
- The system approach
of
multi-modal reasoning and a comparison with other systems. [p89]
- The evaluation goals
of this
system. [p90]
Knowledge-based computerized decision-support
system
applied to the long-term follow-up of bone marrow transplant patients
at the Fred Hutchinson
Cancer Research
Center. [p85]
The system is intended to
To solve a target patient case, this system depends on the cooperation
varied sources
of knowledge and related reasoning types.
Evidence-based reasoning in the context of bone-marrow
transplants (BMT).
[p86]
The quality of practice in medical information systems is described
by two
characteristics upon which knowledge is grounded:
Reliability: Knowledge
generated by
the following and associated with a specific quality of practice, in
decreasing
order of reliability:
world-wide committee of experts (practice principles in a monograph),
committee
of experts (practice guidelines), group of experts (practice
pathways), one
expert (practice cases), one non-expert (medical hokum).
Certainty: The certainty of
knowledge
is related to the proof that validates it. In decreasing order of
certainty:
world-wide controlled clinical trials, controlled clinical trials,
uncontrolled
trials, an individuals experience, no evidence.
The system will provide different access to its knowledge base in
carrying
out clinical tasks, depending on the user's authorization and
expertise levels.
This paper discusses the task of reactive problem-solving. For patient
problem-solving in this system, the knowledge used belongs to practice
guidelines,
practice pathways, and practice cases.
Case-based reasoning, the principle
methodology of the
system. [p.86]
General cycle of case-based reasoning:
A target case to solve
ct
is presented.
Interpretation step Ri
Retrieval step Rt
Reuse step Ru
Revision step Rv
Memorization Rm
A target case ct is a problem to be solved by the
system: ct
= [Si, Sf,
Goal], where Si
is the initial situation, Sf is the final
situation,
and Goal is the result expected from the system.
A memorized case cs
includes a solution in addition to the representation of a target
case.
Rule-based reasoning, based on the RETE
algorithm of
production systems. [p. 87]
The system is also a RBR system in that it contains knowledge in
addition to
cases that are expressed as rules.
The system performs reasoning RBR by firing the elements of a
rule-base to
perform T, a set of cognitive tasks (i.e. diagnosis).
Each rule is of this general format: if condition
then action.
General cycle of rule-based reasoning:
Target problem to solve
ct
is presented.
Interpretation step Ri
Pattern-matching step Rc
Conflict resolution step
Rt
Production step Rp
Update step Rn
The working memory W of the production step contains the
updated
problem description and therefore evolves during the reasoning
process. Updates
do not modify the knowledge base, hence
The action part of a rule contains rule r that is matched during
RBR with Si, then with Sj
the evolution of the working memory and finally until Sf
is satisfied, or Goal if it is empty.
Knowledge representation for various types of
knowledge
within the system. [p. 88]
A network of entities (i.e. practice guidelines)
comprise
the system knowledge base, br
style='mso-special-character:line-break'>
A domain ontology: set of
class
symbols C (concepts) and R. Classes are organized in a
polyhierarchy of classes with several main
categories.
A set of individual symbols
(instances) I, some referring to instances of classes, numbers,
dates,
or other values.
A set of operator symbols
O:
allow logical expressions compsed of
classes,
instances, relationships, and other values; comprised of the
following:
^ (AND), v (OR), ment
guidelines, both defined by the standard practice committee.
The system reasoning. [p.89]
The reasoning merges the reasoning steps of both case-based and
rule-based
reasoning into the multi-modal reasoning cycle.
After the system is presented with a new problem to solve, it is
processed
by the screening step R>The patient-problem solving tasks
follow the following steps:
Interpretation (Ri)
Knowledge search (Rs)
Conflict resolution (Rf)
Reuse (Ru)
Update (Rn)
Memorization (Rm)
The user receives a user-friendly formatting of the entities
retrieved,
differentiating between guidelines, practice pathways and cases.
The system approach of multi-modal reasoning
and a
comparison with other systems. [p. 89]
Systems that achieve the cooperation of case-based reasoning and
other
knowledge representation formalisms (i.e. RBR) can be placed into two
categories:
RBR is the main reasoning
process and
CBR is primarily a heuristic to improve the RBR.
CBR is the main reasoning
process and
RBR or MBR (model-based reasoning) are used "to take advantage of
a
partial domain model available for one part of another reasoning
process."
In this system, it cannot be stated whether CBR or RBR is the main
reasoning
process. It does however attempt to achieve greater cooperation
between the two
by separating the reasoning steps in each of them.
Cognitive Science Summaries Webmaster:
JimDavies (jim@jimdavies.org)
Last modified: Thu Apr 15
11:07:19 EDT
1999