Transcript Computer-Based Tutoring of Medical Procedural Knowledge

Larichev O., Naryzhny Y., Computer-based Tutoring of Medical Procedural Knowledge.
In: Lajoie S., Vivet M. (Eds.) Artificial Intelligence In Education. Open Learning
Environments: New Computational Technologies to Support Learning, Exploration and
Collaboration. 1999, IOS Press, pp. 517-523.
Computer-Based Tutoring of
Medical Procedural Knowledge
Oleg Larichev & Yevgeny Naryzhny
Copyright © 1999, All Rights Reserved
Institute for Systems Analysis of Russian Academy of Sciences
Moscow, Russia
Presented on AI-ED99: 9th International Conference on
Artificial Intelligence in Education, 1999, Le Mans, France
Introduction
Domain
knowledg
e
Declarativ
e
Procedural
(skills)
Changes in Thinking
10 years
Novice
Expert
• Little knowledge
• Rich knowledge
• Poor structure
• Good structure
• Backward reasoning
• Forward reasoning
• Many errors
• Little errors
• Unability to verbalize
Diagnostics as Classification
Operation in anamnesis
Pain in thorax
Suddenly occured dyspnea
Normal arterial pressure
...
aj  A
class ~Cn
class Cn
Operation in anamnesis
Pain in thorax
Suddenly occured dyspnea
Low arterial pressure
ai  A
...
A set of attributes:
P = {P1, P2, …, PM}
Scales of possible values:
Pi = {pi1, pi2, …, piki}
The problem space:
A = P1x P2x …x PM
Problem Space Partial Order
Class Cn
Attribute Pi
ppi3i1
ppi1i1
ppi2i1
ppi4i1
More typical
values for Cn
Less typical
values for Cn
A less typical
object for class Cn
The most typical
object for class Cn
ai  Cn A
Building the Classification
if the expert classifies
this case to class Cn...
...all the more typical
objects belong to class Cn
automatically
ORCLASS Suite
Class Boundary
class Cn
The boundary of class Cn
Discriminative attributes
(pik & … & pik+l)
and at least
t typical values of
attributes {Pr, ...,Pr+s}
Additive attributes
Classification Complexity
class Cn
The most difficult objects
for classification
class ~Cn
Less difficult objects
for classification
Learning Principles
The decision rules are not shown
 Implicit learning via problem solving
 Inductive student model
 Immediate feedback
 Explanations as the expert’s hints
 Gradually increasing complexity
 A big number of tasks

OSTELA Tutoring System
Teaching the Art of Differential Diagnostics
of the Pulmonary Artery Thromboembolism
and the Acute Infarct of Myocardium
Basic Steps: Introduction
Background Knowledge Test
The preliminary knowledge
of typical signs and findings
is a must
Pre-test of Diagnostical Skills
A test case
Instructions for the
learner
Possible diagnoses for selection
Main Learning Step
A training
case
Current
ECG
Possible diagnoses for selection
Feedback on Errors
The case
Expert’s
Hints
The correct diagnosis
OSTELA Training Results
Young physicians of the
Botkin Clinical Hospital
40-60% correct
diagnoses
in the pre-test
About 500 cases
solved
90-100% correct
diagnoses
in the post-test
Unability to
verbalize decision
rules