Rutgers Computers in Biomedicine Project [Rutgers-AIM] Section 9.2.8

Nagel, D. and N.S. Sridharan, (1978) "A Test Procedure for Checking
Combinations of Relational Flags in AIMDS”", Report CBM-79, Department
of Computer Science, Rutgers University, September 1978.

Politakis, P., (1979) "Evolution of an EXPERT Consultation System in
Rheumatology", CBM-TR-99, September, 1979,

"Designing Consistent

Politakis, P., S. Weiss, and C. Kulikowski, (1979)
Systems", CBM-TR-100, September

Knowledges Bases for Eapert Consultation
1979.

$s

Sangster, B.C. (1978) "Natural Language Dialogue with Data Base Systems:
Designing for the Medical Environment", Proceedings of the Third
‘Jerusalem Conference on Information Technology, August: 1975

Schmidt, C.F., and N.S. Sridharan, (1978) Abstract, BELIEVER Project,
report on the Workshop on “Social Plans and Language”, P. Cohen and B.
Bruce (eds.), SIGART Newsletter, No. 6, p. 5, August 1978.

Smith, D. and R. Smith, (1979) "Rules of Student/Teacher Interaction in
Language Acquisition: A Computational Modei”, Report CBM-TR-93, DCS,
Rutgers University, August 1979,

Sridharan, N.S. and Smith, D., (1978) "Design for a Plan Hypothesizer",
Proceedings of the AISB/GI Conference, Hamburg, July 1978; (also, CBM-
TR-85, DCS, Rutgers University).

Sridharan, N.S. and Schmidt, C.F., (1978) "Knowledge-Directed Inference in
BELIEVER", In D. Waterman & F, Hayes-Roth (eds.), Pattern Directed
Inference Systems New York, Academic Press, 1978, pp. 360-379.

Sridharan, N.S. (1978) "Guest Editorial", Special Issue on Applications of
AI to the Sciences and Medicine, Artificial Intelligence Vol 11, No,
2, August 1978.

Sridharan, N.S. and K.N. Venkataraman, "A Planning System Used for Plan
Recognition in a Common Sense Domain”, Report CBM-TM-83, Department of
Computer Science, Rutgers University.

Sridharan, N.S., (1978) "Some Relationships between BELIEVER and TAXMAN",
Report LRP-TR-1, Department of Computer Science, Rutgers University,
December 1978.

Trigoboff, M.L. (1978) "IRIS: A Framework for the Construction of Clinical
Consultation Systems", PhD Dissertation Rutgers University, May 1978;
(also, CBM-TR-90, DCS, Rutgers University)

Van der Mude, A. and A. Walker (1978) "On the Inference of Stochastic
Regular Grammars" Information and Control, (to appear).

Venkataramen, K.N. and N.S. Sridharan, (1978) "An AIMDS Implementation of a

Plan Hypothesizer", Report CBM-TM-80, Department of Computer Science,
Rutgers University, December 1978,

E. A. Feigenbaum 276 Privileged Communication
Section 9.2.8 Rutgers Computers in Biomedicine Project [Rutgers-AIM]

Weiss, S.M., Kern, K.B., Kulikowski, C.A. and Pincus, W., (1978) "An
Interactive System for the Design of Classifiers in Diagnostic
Applications", Proc. International Conference on Cybernetics and
Society, Tokyo, November 1978.

Weiss, S.M., Kulikowski, C.A., and Safir, A., (1978) "Glaucoma Consultation
by Computer", Computers in Biology and Medicine, 8:25-40, January 1978.

Weiss, S.M., Kulikowski, C.A., Amarel, S., and Safir, A. (1978) "A Model-
Based Method for Computer-Aided Medical Decision Making", Artificial
Intelligence Special Issue on Applications to the Sciences and Medicine
(11, 1978); also, CBM-TR-63, DCS Rutgers University April 1978).

Weiss, S.M., Kulikowski, C.A., Mizoguchi, F. and Kitazawa, V. (1978), "A
Computer-Based Comparison of Japanese and American Decision-Making"
Proc. International Conference on Cybernetics and Society, Tokyo,
November 1978.

Weiss, S., (1979) "The EXPERT and CASNET Consultation Systems", Proc.
Japan-AIM Workshop, Tokyo, August 1979,

Weiss, S. and C. Kulikowski, (1979) "EXPERT: A System for Developing
Consultation Models", Proc. IJCAI, Tokyo August 1979, pp. 942-947,
[also Rutgers Computer Science Report CBM-TR-97].

Weiss, S., C. Kulikowski, and B. Nudel (1979) “Learning Production Rules
for Consultation Systems, Proc. IJCAI, Tokyo, August 1979, pp 948-950.

Weiss, S., K. Kern, and C. Kulikowski, (1978) "A Guide to the Use of the
EXPERT Consultation System", CBM-TR-94, November 1978.

E) Funding Support

The Rutgers Resource is funded through an NIH grant: Research
Resource on Computers in Biomedicine. The NIH grant number is P 41 RR643.
The Director and Principal Investigator is Dr. Saul Amarel of Rutgers--The
State University of New Jersey. .

This grant is in its third renewal extending for three years from
December 1977 through November 1980. The total amount of the award for
this period is $1,426,598 in direct costs. In the current year, December
1979 through November 1980, the funding level of direct costs is $451,383.

Il. INTERACTIONS WITH THE SUMEX-AIM RESOURCE

A. Medical Collaborations and Dissemination: Interactions

The SUMEX-AIM facility provides one of the nodes where a good part of
our collaborative program development and testing takes place (the other

facility is RUTGERS/LCSR). These medical collaborations are described in
I.B. above.

Privileged Communication 277 E. A. Feigenbaum
Rutgers Computers in Biomedicine Project [Rutgers-AIM] Section 9.2.8

An important responsibility of the Rutgers Resource within the
National AIM community is to sponsor dissemination and training activities.
The focus of our efforts in this area continues to be centered around the
AIM Workshops and sessions on AIM research at national and international
conferences.

As part of our collaborative activities with SUMEX-AIM in this area,
we have continued our contribution to the preparation of the AI Handbook,

In order to increase the dissemination of AIM work within specialty
fields of medicine, we have also presented tutorial papers at relevant
conferences.

1) Fifth AIM Workshop (1979)

The fifth Annual AIM Workshop differed from the ones preceding it.
It was a mini-workshop devoted to a single sub-area of AIM research:
medical consultation problems and systems. This year, responsibility for
the organization of the miniworkshop rotated to Dr. Peter Szolovits of MIT,
with the intention that a system of rotation for the hosting will now
develop (next year's will be held at Stanford). The Rutgers Resource
retained a coordinating and funding role: Dr. Kulikowski worked with Dr.
Szolovits in the organization of the miniworkshop, and the Resource covered
the travel expenses of a number of AIM groups and individual participants.

The miniworkshop took place at the Talbot House, S. Pomfret, Vermont
on May 7-20, 1979, and attendance was limited by a number of people that
could be accommodated. A deliberate attempt was made to include only those
investigators from medicine and computer science involved in the day-to-day
activities of AIM research. Participation by graduate students was
encouraged. there were 32 attendees, including 11 computer scientists, 14
graduate students and 7 physicians. In contrast to previous workshops, the
concentration on a single research area allowed greater depth of
discussion, and encouraged more informal interchange of ideas and opinions.
It also served an important training function for graduate students and
junior AIM researchers, who were not able to attend IJCAI because of its
distant location (Tokyo) this year.

The general structure of the miniworkshop was as follows:

a) Five technical sessions on computer science issues of the
representation of knowledge, explanation and justification, knowledge base
acquisition and maintenance, and the computational methods of INTERNIST-11.

b) Three general AIM sessions on the assessment of the current state
of progress of AIM methods and programs; the sources of difficulties in
bringing these programs to clinical application; and an experimental
discussion of knowledge-acquisition problems centered around a simulated
CPC-style protocol analysis.

The miniworkshop format was successful in encouraging a more intimate

exchange of ideas among workers in this AIM subfield. It would be
desirable to hold other miniworkshops in the future in application areas

E. A. Feigenbaum 278 Privileged Communication
Section 9.2.8 Rutgers Computers in Biomedicine Project [Rutgers-AIM]

such as psychology and biochemistry as the interest of the AIM researchers
dictates. The format of a general workshop for all groups should probably
be reserved for less frequent meetings.

2) AIM-Japan Workshop:

A one-day workshop was organized by Drs. Kaminuma, Kurashina, Kaihara
and Mizoguchi to follow the IJCAI meeting in Tokyo. It was held on August
25, 1979, and was attended by over 100 biomedical researchers and clinical
investigators. The workshop consisted of presentations by U.S. AIM
researcners and Japanese investigators - two of them collaborators of the
Resource,

Prior to the presentations, Drs. Kulikowski and Weiss gave
demonstrations of the CASNET/Glaucoma and EXPERT/Rheumatology programs and
Dr. Shortliffe demonstrated the MYCIN system. These were available for
testing and study by the participants. In addition, Dr. Mizoguchi
demonstrated the EXPERT/Giaucoma program that incorporates in its knowledge
base a core of CASNET/Glaucoma and the new structures introduced by Dr.
Kitazawa. The AIM-Japan Workshop had the effect of disseminating knowledge
of AIM research among a large group of Japanese clinical investigators, and
the attendance of a number of other IJCAI attendees lent an international
character to the meeting.

3) IJCAT - International Joint Conference on Artificial
Intelligence (Resource Participation):

Included among the other AIM activities that took place at the
conference, Drs. Amarel and Mitchell] chaired sessions, Drs. Kulikowski and
Weiss presented two joint papers, and a third one was presented by our
ophthalmology collaborators in Japan, Drs Kitazawa and Mizoguchi, who
presented results of their experiments with the EXPERT knowledge-based
methods. Drs. Kulikowski and Weiss demonstrated the CASNET/Glaucoma,
EXPERT/Rheumatology and EXPERT/Thyroid consultation programs during the
special demonstration periods at the conference. We connected by TYMNET
via satellite to the Rutgers/LCSR DEC-20 and had excellent response time.
This demonstrated the feasibility of remote collaborations and sharing of
resources for the future. Dr. Mizoguchi demonstrated the EXPERT/Glaucoma
program using Dr. Kitazawa's knowledge base, on the FUJIMIC DEC-20 in
Tokyo, illustrating the practicality of knowledge base transfer methods.

4) National AIM Projects at Rutgers

The national AIM projects approved by the SUMEX-AIM executive
committee were increased during the 1979 period of th Resource. A project
using the BRIGHT system developed within the Resource and the NIH was
continued in its application to various problems of clinical research by
the group headed by Dr. W. Gordon Walker at the Johns Hopkins University
and Hospital. We have projects that have their primary locus of activities
on the SUMEX system and also use the Rutgers Resource for development,
testing, and back-up functions. These include the MAINSAIL and CONGEN
projects from Stanford University, and the INTERNIST project at the
University of Pittsburgh. A project on medical knowledge representations

Privileged Communication 279 E. A. Feigenbaum
Rutgers Computers in Biomedicine Project [Rutgers-AIM] Section 9,2.8

at the Ohio State University was initiated on the Rutgers Resource
Computer, as was a project in Artificial Intelligence models of clinical
reasoning developed by Dr. R. Greenes at Harvard University. Dr. David
Garfinkel at the University of Pennsylvania developed programs for his
project on metabolic pathway modeling on the Resource Computer.

B) Critique of SUMEX-AIM Resource Management

We have now reached a steady state level of SUMEX-AIM usage--at least
for the foreseeable future; we estimate it to remain at about 750 connect
hours per year with an average compute to connect ratio of 1:25.

Since December 1979, the RUTGERS/LCSR computer is connected to the
ARPANET again. Access to the ARPANET is facilitating close interactions
between the Rutgers and Stanford AIM facilities, and in particular between
the system staffs.

We continue to find the people support at SUMEX-AIM first rate and

extremely helpful. On the technical side, we find communications via

TYMNET of questionable reliability; and the SUMEX-AIM computer too heavily
loaded,

TIT. RESEARCH PLANS

A) Project Goals and Plans

 

We are planning to continue along the main lines of research that we
have established in the Resource to date. Our medical collaborations will
continue with emphasis on development of consultation systems in
rheumatology and ophthalmology. Work on belief systems and commonsense
reasoning will continue with emphasis on the psychology of plan recognition
and handling of stereotypes. Our core work will continue with emphasis on
further development of the EXPERT framework and also on AI studies in
representations and problems of knowledge and expertise acquisition. We
also plan to continue our participation in AIM dissemination and training
activities as well as our contribution--via the RUTGERS/LCSR computer--to
the shared computing facilities of the national AIM network.

In October 1979 we submitted to NIH a renewal proposal for the
Rutgers Resource. Our proposal for a five-year continuation (December 1,
i980 to November 30, 1985) was reviewed by a special study section earlier
this April. A decision by NIH is expected in late May.

B) Justification and Requirements for Continued SUMEX USE

 

Continued access to SUMEX is needed for:
1} Backup for DEMOS, etc.
2) Programs developed to serve the National AIM

Community should be runnable on both facilities,

E. A. Feigenbaum 280 Privileged Communication
Section 9.2.8 Rutgers Computers in Biomedicine Project [Rutgers-AIM]

3) There should be joint development activities
between the staffs at Rutgers and Sumex in
order to ensure portability, share the load,
and provide a wider variety of inputs for
developments.

C) Needs and Plans for other Computing Resources Beyond SUMEX-AIM

Beyond the current SUMEX-AIM facility there is need for access toa
more ‘personal’ type computing facilities (e.g. PERQ, DORADO, LISP
machines, etc.) In addition SUMEX might provide a high quality output
device (e.g. line printer or XGP) for the community.

D, Recommendations for Future Community and Resource Development

Future development for hardware should be in the direction of smaller
machines which could ultimately ve acquired at or transferred to user's
sites (e.g. VAXes or the larger personal computers). Special efforts in
networking small machines and in developing methods of using small
computers would be desirable. In particular, methods and technology for
System transfer from large machine environments to small machines would be
increasingly useful to the AIM community.

We continue to consider community developments as one of the
Significant goals of the national AIM project. The program of AIM
Workshops should continue and new arrangements involving a program of
lectures/seminars and working visits by AIM scientists should be
encouraged.

Privileged Communication 281 E. A. Feigenbaum
Decision Models in Clinical Diagnosis [Rutgers-AIM] Section 9.2.9

9.2.9 Decision Models in Clinical Diagnosis [Rutgers-AIM]

 

A Goal-Oriented Model of Clinical Decision-Making
Incorporating Decision Thresholds

Robert A. Greenes, M.0., Ph.D.
Harvard Medical School
Department of Radiology

Peter Bent Brigham Hospital
Boston, Massachusetts

I. Summary of Research Program
A. Project Rationale

The major objective of this project is to increase understanding of
the way in which probability of diagnosis, and costs and benefits of
contemplated actions, interact in the selection of the most appropriate
actions. Actions include therapeutic as well as diagnostic procedures.
The initial problem area being considered is the management of a patient
with upper abdominal pain. The decision problem is modeled as a goal-
oriented search process, where actions available for selection represent
the system's goals. Costs and benefits of the actions are incorporated in
the heuristic device of a probability threshold, which must be exceeded for
the action to be taken. Production rules, modified to incorporate
evaluation of probabilities in relation to thresholds, are used to embody
the decision conditions.

Top-level patient management goals usually require for their adoption
that diagnostic thresholds be exceeded, which in turn require specific
groups of diagnostic tests to be obtained. Selection of these tests may
recursively require still other diagnostic thresholds to be exceeded,
involving other preliminary diagnostic tests. A forward-chaining search
Strategy is used, which restricts consideration to diagnoses exceeding
minimal “rule-out" thresholds. When a group of diagnostic tests is
eligible to be done, based on the above search, its members are placed in
an eligibility list. Concurrent search may find several eligible test
groups. Selection among eligible tests is based on heuristics such as:
greatest likelihood of enabling a higher level goal to be reached, highest
number of goals to which the test relates, and least overall cost or risk.

At the initiation of a session, information already known about a
patient is entered. This, together with the system's estimates of
prevalence of the various diseases considered, is used to "prime" the
differential diagnosis probability distribution by means of Bayes theorem.
The system identifies a next test to perform. As results on tests become
available and are entered into the system, Bayes theorem is again used to
update the distribution.

E. A. Feigenbaum 282 Privileged Communication
Section 9.2.9 Decision Models in Clinical Diagnosis [Rutgers-AIM]

B. Medical Relevance and Collaboration

This model is somewhat unique in its attempt to incorporate both the
decision analytic view of costs and benefits, and heuristics which make the
problem more tractable. A probability threshold may be viewed as the
indifference point for a decision maker, where the relationship between
cost and benefit is equal for any of the available options. It can thus be
used as a device for collapsing and summarizing an entire distal decision
tree.

By maintaining the differential diagnosis in probabilistic terms,
rather than using various surrogates for probability, the model is able to
relate diagnostic probability to action, through the translation provided
by the threshold concept. This permits behavior of the model to be
analyzed and tuned either in terms of the accuracy of its probability
estimates or the suitability of thresholds. Further, we believe that the
use of a "rule-out” threshold, to permit the model to focus on limited
Subsets of the possible diagnoses, bears a resemblance to the focusing
process which medical problem solving actually exhibits.

This project involves internal collaboration between the Departments
of Radiology and Medicine at Peter Bent Brigham Hospital, with consultation
by the Department of Biostatistics at Harvard School of Public Health, Bolt
Beranek and Newman, Inc., and the Office of Medical Education, Michigan
State University.

Cc. Highlights of Research Progress

In this first year of effort, major tasks have involved: (1)
development of prototype programs for acquisition of decision rules,
construction of probability tables, output of rules and probability data,
and execution of the decision-making system; and (2) elaboration of
Specific decision rules and probability estimates for the management of
patients with upper abdominal pain. An operational prototype of each of
the programs described above now exists. In the application to abdominal
pain, we consider approximately 90 diagnostic entities, 40 management
goals, and 60 individual tests. Thus far, we have concentrated on the
subset of patients suspected of having gastrointestinal obstruction,
involving 6 diagnostic entities. The rules and probabilities are derived

subjectively by periodic sessions with a gastroenterologist, T. E. Bynum,
M.D.

D. Relevant Publications

. [1] Greenes RA: A goal-directed model for investigation of thresholds
for medical action. Proc Sympos on Computer Applications in Medical Care,
Washington, DC, October, 1979, IEEE, pp 47-51.

[2] Greenes RA, McNeil BJ: The use of statistical measures as an aid
to selection of appropriate diagnostic procedures. Proc 65th Scientific
Assembly, RSNA, Atlanta, GA, November, 1979, p 257 (abstract).

Privileged Communication 283 E. A. Feigenbaum
Decision Models in Clinical Diagnosis [Rutgers-AIM] Section 9.2.9

[3] Greenes RA: The diagnostic test order decision problem. Proc of
the 6th Illinois Conf on Medical Information Systems, Champaign-Urbana, 11,
April, 1980 (in press).

[4] Greenes RA: Medical decision-making research: the role of
academic radiology. Third Int Sympos on the Planning of Radiological
Departments. Amsterdam, Holland, June, 1980.

E. Funding Support

This project is part of a Program Project, "Investigations in
Clinical Decision-Making", supported by the National Library of Medicine,
grant NLM 1 PO1 LM03401, Robert A. Greenes, M.D., principal investigator.
Total award (7/1/79-6/30/84), $1,177,582. First year (7/1/79-6/30/80),
$235 , 582,

II, Interactions with the SUMEX-AIM Resource
A. Medical Collaboration and Program Dissemination Via SUMEX

With this project only in its early stages, no significant medical
collaboration or dissemination has yet occurred. Because of the specific
features of our model, it was not considered to be readily implemented
within the framework of other extant decision-making systems.

B. Sharing and Interactions with Other SUMEX-AIM Projects

The project utilizes the PDP-20 AIM resource at Rutgers University.
In making the decision to utilize the Rutgers rather than the SUMEX
facility, much assistance and documentation was provided by the technical
directors of both facilities. Our choice of Rutgers was based primarily on
the expectation that the response time and communication support through
TYMNET for an east-coast user would be likely to be better.

We have participated in the site visit to the AIM resource at Rutgers
in April, 1980, and will also be participating in the AIM Workshop, and
Artificial Intelligence in Medicine Continuing Education Tutorial, at
Stanford, in August, 1980.

C. Critique of Resource Management

We are most pleased by the personal interest shown, and assistance
provided, by the technical directors of both AIM resources. We have had no
serious problems with the use of the Rutgers facility.

III. Research Plans (8/80-7/86)

A, Project Goals

Near term goals involve (a) expansion of the decision rules and
probability matrix to include the entire range of diagnostic entities
considered in our model of upper abdominal pain, (b) incorporation of the
time duration involved in diagnostic tests into the selection process, and

E. A. Feigenbaum 284 Privileged Communication
Section 9.2.9 Decision Models in Clinical Diagnosis [Rutgers-AIM]

(c) evaluation of model performance. Initially, our criteria for
evaluation will be agreement with an expert regarding the management goals
selected, and the tests utilized.

Long-range goals include (a) refinement of the human interaction with
the system, (b) the incorporation of capabilities for explaining its
decisions, (c) evaluation of sensitivity of its conclusions to estimates of
the various probabitities and thresholds, and (d) incorporation of
empirical probability data into the model when available. Ultimately, wa
would Tike to evaluate its suitability as a consultant.

B. Justification and Requirements for Continued SUMEX Use

We expect the complexity of our programs to grow considerably during
the next 2-3 years. The knowledge and data bases will grow also, and we
anticipate moderately large storage requirements. Continued availability
of the SUMEX-AIM resource is thus highly desirable in terms of our need for
LISP programming capabilities.

In addition, we would hope that closer interaction with other AIM
users in the evaluation of our model and other approaches will become
possible, as we hoth familiarize ourselves with the characteristics of
other systems, and further develop the capabilities of our system.

C. Needs and Plans for Other Computing Resources Beyond SUMEX-AIM.

This project has no present need for other computing resources,
although some of the probability data incorporated into the model are
derived from studies carried out on other computer systems.

D. Recommendations for Future Community and Resources Development

As microcomputers become increasingly powerful, inexpensive, and
capable of supporting at least single-user LISP programs, wa expect that a
natural evolution toward such systems will occur. Efforts to ensure
compatibility, portability, and the ability to interface such systems to
the AIM network will, thus, be highly desirable.

Privileged Communication 285 E. A. Feigenbaum
Heuristic Decisions in Metabolic Modeling [Rutgers-AIM] Section 9.2.10
9.2.10 Heuristic Decisions in Metabolic Modeling [Rutgers-AIM]

Heuristic Decisions in Metabolic Modeling

David Garfinkel, Ph.D,

Moore School of Electrical Engineering
Department of Computer and Information Science
University of Pennsylvania
Philadelphia, Pennsylvania

T. SUMMARY OF RESEARCH PROGRAM

 

A. Project Rationale

This research is concerned with developing methods of constructing
computer models of complex metabolic systems, and applying thereto
artificial intelligence and other relevant computer science techniquas.

B. Medical Relevance

Most of our work is concerned with modeling cardiac metabolism and
the effects of ischemic heart disease. There has been some collaboration.
with appropriate cardiac physiologists and cardiologists. We are trying to
extend our reseaich to diabetes and hematology, in collaboration with
experts in those particular fields.

C. Research Progress

We constructed cardiac metabolism models, emphasizing the effects of
acidosis, fatty acid metabolism, and the glycogenolysis cascade; this work
is still in progress. We have developed methods for sensitivity analysis
and a new program for building these models. Current efforts emphasize
completing the glycogenolysis model, correcting some of the existing
models, designing relevant experiments, and writing up a mass of completed
models. We are extending our model-building software, and making it
"friendly" enough for unsophisticated users.

D. List of Relevant Publications

M.J. Achs and D. Garfinkel, Metabolism of the Acutely Ischemic Dog Heart.
I. Construction of a Computer Model. Am. J. Physiol. 236, R21-R30
(1979).

D. Garfinkel and M.J. Achs, Metabolism of the Acutely Ischemic Dog Heart.
Ii. Interpretation of a Model. Am. J. Physiol. 236, R31-R39 (1979).

D. Garfinkel, M.J. Achs, M.C. Kohn, and L.S. Menten, Modeling of Complex

Metabolic Systems as a Composite of Standard Computer Science
Techniques, Proceedings of Summer Computer Simulation conference, 1979.,.

—E. A. Feigenbaum 286 Privileged Communication
Section 9.2.10 Heuristic Decisions in Metabolic Modeling [Rutgers-AIM]

L. Garfinkel, M.C. Kohn, and D. Garfinkel, Computer Simulation of the
Fructose Bisphosphatase-Phosphofructokinase Couple in Rat Liver, Eur.
J. Biochem. 96, 183-192 (1979).

M.C. Kohn, L.E. Menten, ad D. Garfinkel, A Convenient Computer Program for
Fitting Enzymatic Rate Laws to Steady-State Data, Comput. Biomed. Res.
12, 461-469 (1979).

M.C. Kohn, M.J. Achs and D. Garfinkel, Computer Simulation of Metabolism in
the Pyruvate-Perfused Rat Heart. I. Model Construction. Am. J. Physiol.
237, R153-158 (1979).

L.E. Menten, M.C. Kohn and D. Garfinkel, A Convenient Computer Program for
Estimation of Enzyme and Metabolite Concentrations in Multienzyme
Systems (in progress).

E. Funding Support

1.) “Computer Simulation in Cardiology” HL 15622, 3 years, Dec. 1,
1977-Nov. 30, 1980; current year $111,051. Competing renewal now pending,
$162,744; $181.060; $199,548; $220,647 (for four years).

2.) "Computer Modeling Methods for Metabolic Systems," (GM 16501-
11A1), $60,598; $63,860 (two years) April 1, 1980 - March 31, 1982.

3.) "Computer-Aided Study of Glycolysis in Pancreatic Islets,"
submitted to NIH as a part of the Diabetes Center renewal proposal. Direct
costs $38,853; $41,652; $44,322 for 3 years.

4.) "Metabolism of Maltarial, Aged, and Normal Erythrocytes"
(including experimental subcontract) submitted to NIH, $144,283; $158,267;
$169,609 (3 years).

II. Interactions with the SUMEX-AIM Resource

There has been no interaction with SUMEX directly, and none is
anticipated. There has been considerable interaction with the Rutgers
resource, especially in the form of a personal collaboration with Prof.
Kulikowski, and moderate usage of their computer. The program we have
developed and are writing up was developed there. I have also been a
regular attendant at AIM workshops over the last few years, and have made
valuable contacts and acquired an understanding of the field that would not
have been possible otherwise. Overall, interactions through the Rutgers
resource have been of considerable importance to our development.

III. Research Plans

We are now at the stage where our technique research is concerned
with knowledge representation and acquisition, with intelligent (and fuzzy)
data bases having some of the characteristics of a knowledge base, with
representation of knowledge in our subject area of interest, and with
overcoming incompleteness. We cannot realistically expect much help in
these matters from the subject-matter experts, and need the help of the AI

Privileged Communication 287 E. A. Feigenbaum
Heuristic Decisions in Metabolic Modeling [Rutgers-AIM] Section 9.2.10

community. We expect to continue using the Rutgers.machine. Within the
ext year and a half we will try to refine our existing program and other
model-building methods, to make them more manageable, modularized,
efficient, and faster, and also friendly enough to be operated by
biological experts directly without too much help from programmers or
professional simulationists. We will also devise methods (using symbolic
manipulation) to break large complex models down into workably small
pieces. In the next few months we hope to start attacking the data-base
(and Tater knowledge-base) aspects of this work.

Long-range research goals will be critically dependent on other
funding, so we cannot give details now. We hope to be able to build a good
model of cardiac ischemia which can be used to make predictions, to design
experiments for areas in metabolism (especially multi-enzyme systems) which
are now inefficient. This modeling process must be fast enough to be of
use and rigorous enough to be reliable. This implies development of the
techniques mentioned above to the point where they can do most of the
necessary work.

We do not expect to use the SUMEX computer, but do expect to make
considerable use of the Rutgers computer and the expertise of the
department, since such expertise is not otherwise available to us. It is
conceivable that several years from now we may want to link in a personal
computer under the collaborative linkage described in the RENEWAL
RATIONALE, but this is sufficiently far in the future that a justification
for one cannot be given at the present time.

E. A. Feigenbaum 288 Privileged Communication
Section 9.3 Pitot Stanford Projects

9.3 Pilot Stanford Projects
The following are descriptions of the informal pilot projects

currently using the Stanford portion of the SUMEX-AIM resource pending
funding, and full review and authorization.

Privileged Communication 289 E. A. Feigenbaum
Ultrasonic Imaging Project Section 9.3.1
9.3.1 Ultrasonic Imaging Project

Ultrasonic Imaging Project

James F. Brinkley, M.D.
W.D. McCallum, M.D.
Depts. Computer Science, Obstetrics and Gynecology
Stanford University

I. Summary of Research Program
A. Project Rationale

The long range goal of this project is the development of an
ultrasonic imaging and display system for three-dimensional modeling of
body organs. The models will be used for non-invasive study of anatomic
Structure and shape as well as for calculation of accurate organ volumes
for use in clinical diagnosis. Initially, the system will be used to
determine fetal volume as an indicator of fetal weight; later it will be
adapted to measure left ventricular volume, or liver and kidney volume.

The general method we plan to use is the reconstruction of an organ
from a series of ultrasonic cross-sections taken in an arbitrary fashion. A
real-time ultrasonic scanner will be coupled to a three-dimensional
acoustic position locating system so that the three-dimensional orientation
of the scan plane is known at all times. During the patient exam a
dedicated microcomputer based data acquisition system will be used to
record a series of scans over the organ being modelled. The scans will be
recorded on a video disk which is controllable by the microcomputer. 3D
position information will be stored on a floppy disk file. The
microprocessor wil? then be connected to SUMEX where it will become a slave
to an AI program running on SUMEX. The SUMEX program will use a model
appropriate for the organ which will form the basis of an initial
hypothesis about the shape of the organ. This hypothesis will be refined at
first by asking the user relevant clinical questions such as (for the
fetus) the gestational age, the lie of the fetus in the abdomen and
complicating medical factors. This kind of information is the same as that
used by the clinician before he even places the scan head on the patient.
The model will then be used to request those scans from the video disk
which have the best chance of giving useful information. Heuristics based
on the protocols used by clinicians during an exam will be incorporated
since clinicians tend to collect scans in a manner which gives the most
information about the organ. For each requested scan a prototype outline
derived from the model will be sent to the microcomputer. The requested
scan will be retrieved from the video disk, digitized into a frame buffer,
and the prototype used to direct a border recognition process that will
determine the organ outline on the scan. The resulting outline will be sent
to SUMEX where it will be used to update the model. The scan requesting
process will then be continued until it is judged that enough information
has been collected. The final model will then be used to determine volume
and other quantitative parameters, and will be displayed in three
dimensions.

E. A. Feigenbaum 290 Privileged Communication
Section 9.3.1 Ultrasonic Imaging Project

We believe that this hypothesize verify method is similar to that
used by clinicians when they perform an ultrasound exam. An initial model,
based on clinical evidence and past experience, is present in the
clinician's mind even before he begins the exam. During the exam this model
is updated by collecting scans in a very specific manner which is known to
provide the maximum amount of information. By building an ultrasound
imaging system which closely resembles the way a physician thinks we hope
to not only provide a useful diagnostic tool but also to explore very
fundamental questions about the way people see.

We plan to develop this system in phases, starting with an earlier
version developed at the University of Washington. During the first phase
the previous system will be adapted and extended to run in the SUMEX
environment. A clinical study will then be carried out to determine its
effectiveness in predicting fetal weight. At the same time computer vision
techniques will be used to develop the system further in the direction of
increased applicability and ease of use. We thus hope to develop a limited
system in order to demonstrate the feasibility of the technique, and then
to gradually extend it with more complex computer processing techniques, to
the point where it becomes a useful clinical tool.

B. Medical relevance

This project is being developed in collaboration with the Ultrasound
Division of the Department of Obstetrics at Stanford, of which W.D.
McCallum is the head.

Fetal weight is known to be a strong indicator of fetal well-being:
small babies generally do more poorly than larger ones. In addition, the
rate of growth is an important indicator: fetuses which are "small-for-
dates" tend to have higher morbidity and mortality. It is thought that
these small-for-dates fetuses may be suffering from placental
insufficiency, so that if the diagnosis could be made soon enough early
delivery might prevent some of the complications. In addition such growth
curves would aid in understanding the normal physiology of the fetus.
Several attempts have been made to use ultrasound for predicting fetal
weight since ultrasound is painless, noninvasive, and apparently risk-free.
These techniques generally use one or two measurements such as abdominal
circumference or biparietal diameter in a multiple regression against
weight. We recently studied several of these methods and concluded that the
most accurate were about +/-200 gms/kg, which is not accurate enough for
adequate growth curves (the fetus grows about 200 gms/week). The method we
are proposing is based on the assumption that fetal weight is directly
related to volume since the density of fetal tissue is nearly constant. We
are hoping that by utilizing three dimensional information more accurate
volumes and hence weights can be obtained.

In addition to its use in predicting fetal weight, this system could
be used to determine other organ volumes such as that of the left
ventricle. Left ventricular volumes are routinely obtained by means of
cardiac catheterization in order to help characterize left ventricular
function. Attempts to determine ventricular volume using one or two
dimensional information from ultrasound has not as yet demonstrated the

Privileged Communication 291 E. A. Feigenbaum
Ultrasonic Imaging Project Section 9.3.1

accuracy of angiography. Therefore, three-dimensional information should
provide a more accurate means of non-invasively assessing the state of the
Teft ventricle.

C. Highlights of Research Progress

During the past year we have essentially completed the first phase of
this project which was to implement and adapt the previous system to the
SUMEX environment. The accomplishments related to that goal are:

1. Completion of a microprocessor based data acquisition system

The following hardware has been obtained and integrated into the
system--

a) A Toshiba real-time ultrasonic phased array scanner, in routine
clinical use at the Dept. of Obstetrics.

b) A Sony video tape recorder and Hitachi monitor, for use in recording
the scans prior to their being outlined with the Tight pen.

c) A custom built acoustic position locating system for determining the
position of the scan plane in space, supplied to us by W.E. Moritz.
at the University of Washington.

d) A Datamedia computer terminal for communicating with SUMEX and
controlling the procedure.

e) A microprocessor-based video graphics system supplied to us by
Varian Corporation. This system includes a light pen, dual floppy
disks and video display memory.

A large amount of software for the data acquisition system has been
written and is now working. This software consists of routines to direct
the patient exam, during which time scans are recorded on video tape and
position information is stored on floppy disk. Additional software directs
that the scans be outlined with the light pen (not digitized in this first
phase) and stored with the position information. Finally, a program has
been written which converts the microprocessor into a video graphics
terminal, Characters are passed back on forth as with an ordinary terminal
but special command sequences cause graphics to be displayed and a file
transfer to take place. The file transfer, called File Transfer to Micro,
is packet oriented and should prove useful to anyone wanting to do file
transfers between SUMEX and a microcomputer. It is also flexible enough so
that it can form the basis of a system for sending commands from SUMEX to
do local image processing functions.

2, Completion of SUMEX high level routines
The SUMEX software for this first phase includes procedures to
transfer the data from floppy disk to SUMEX (via the file transfer

protocol), to build a 3D reconstruction using simple interpolation, and to
display the images on a graphics terminal.

E. A. Feigenbaum 292 Privileged Communication
Section 9.3.1 Ultrasonic Imaging Project

3. Initial tests and first patients

The completed first phase system has been used to build 3D displays
of simple models such as cylinders in a water tank. We have also tried it
on 2 patients and have obtained 3D plots of the fetal head and trunk and of
the placenta inside the uterus.

The research currently in progress relates to testing the system:

1, An engineering study is being carried out on cylinders, balloons
and point targets to determine the bench accuracy of the system.

2. A clinical protocol is being established on several obstetrics
patients. Once we have gained enough experience we will begin our clinical
Study to determine the ability of the method to predict fetal volume and
weight.

D. Publications

Brinkley, J.F., Moritz, W.E., Baker, D.W., “Ultrasonic Three-Dimensional
Imaging and Votume From a Series of Arbitrary Sector Scans", Ultrasound
in Medicine and Biology, vol 4, pp 317-327.

Brinkley, J.F., McCallum, W.D., Daigle, R.E., "A Distributed Computer
System for Fetal Weight Determination", Proceedings of the 24th Annual
Meeting of the American Institute of Ultrasound in Medicine, Montreal,
August 27-31, 1979, p 113.

McCallum, W.D., Brinkley, J.F., "Estimation of Fetal Weight from Ultrasonic
Measurements", American Journal of Obstetrics and Gynecology, 133:2,
pp.195-200, Jan. 1979,

E. Funding status

"Ultrasonic Measurement of Fetal Volume and Weight"
Principal Investigator: W.D. McCallum, M.D.
Assistant Professor
Department of Obstetrics and Gynecology
Stanford University
Funding agency: National Institute of Child Health and Human Development
Number: 1-RO1 HD12327-01
Total term and direct cost: 7/1/79-6/30/81, $111,823
Current funding period: 7/1/79-6/30/80, $60,423

II. Interactions with SUMEX-AIM resource

A. Collaborations

We are collaborating more with medical people than anyone else. The
project is located in the Obstetrics Department at Stanford where W.D.
McCallum manages the ultrasound patients. We have also been discussing the

the applicability of the current system to the heart with Dr. Richard Popp
in the Division of Cardiology at Stanford.

Privileged Communication 293 E. A. Feigenbaum
Ultrasonic Imaging Project Section 9.3.1

B. Sharing and Interactions with SUMEX projects

Mostly personal contacts with the Heuristic Programming Project and
MYCIN project at Stanford. The message facilities of SUMEX have been
especially useful for maintaining these contacts. Since the first phase of
the project is now essentially completed we expect to interact much more
with other SUMEX projects in order to develop the AI ideas.

C. Resource management

In general SUMEX has been a very usable system, and the staff has
been very helpful. The only complaint is that it is impossible to get
anything done in the afternoons since we always get bumped.

IIi. Research Plans
A. Project goals and plans

As mentioned in Part I we plan to implement this system in phases,
each phase requiring use of more sophisticated artificial intelligence
techniques. The major phases arte as follows (in chronological order:

1..Set up prototype system and test its ability to predict
fetal weight.

This system has been developed and is now undergoing testing. We plan
to carry out engineering and clinical studies in order to test the ability
of the current system to predict fetal and cardiac volume. If successful
the system may have clinical impact as it stands. However, our initial
patient studies have demonstrated the basic limitations of the system,
which are inadequate models and difficulty of use. From a medical point of
view the next phases will be attempts to remove these limitations.

2. Explore other methods for geometric modelling, AI techniques
of goal directed problem solving.

In order to develop adequate models and control strategy it will be
necessary to examine other AI methods of generating models and using them
to guide problem solving programs. For this aspect of our research the
SUMEX-AIM community should be especially useful.

3. Develop program, as outlined in the introduction,
with several limitations--

Only a simple organ will be modelled at first, i.e. not the entire
fetus including limbs the computer will still request certain scans to be
retrieved from the video disk but the operator will outline them with the
light pen. Since ultrasound image quality is improving so rapidly it makes
Sense to wait as long as possible before attempting automated border
recognition. The models and control strategies developed during this phase
Should be useful when actual border recognition is attempted however.

E. A. Feigenbaum 294 Privileged Communication
Section 9.3.1 Ultrasonic Imaging Project

4, Extend the technique to more irregular objects structured
models will be developed so that the fetal limbs can be
included.

5. Add image processing hardware, develop automated border
recognition software.

The models developed in the last two phases will be used to guide the
border recognition process.

As these phases are implemented they will continue to be tested
against the clinical data acquired and stored on floppy disk by the data
acquisition system. In this way we can develop new ideas while continually
upgrading the clinical utility of the system.

B. Justification for continued use of SUMEX

The goals of this project seem to be compatible with the general
goals of SUMEX, ie to develop the uses of artificial intelligence in
medicine. The problem of three-dimensional modelling is a very general one
which is probably at the very heart of our ability to see. By developing a
medical imaging system that models the way clinicians approach a patient we
should not only develop a useful clinical tool but also explore some very
fundamental problems in AI.

C. Need for resources
1.SUMEX resources

The only additional requirements we have at present are for an
additional file directory and for a little more time in the afternoon. At
present we only have one directory which must be shared by the system
developer and an additional person conducting the engineering and clinical
Studies. An additional directory could be designated for users of the
current, implementation of the system while the present directory could be
used for new developments.

2. Other resources

Judging from our presént experience it appears that SUMEX could not
handle the amount of data required for image processing on digitized
ultrasound scans. This is one of the main reasons we are proposing a
distributed system in which SUMEX only directs a smaller machine to do the
actual number crunching. It is also one of the reasons we are postponing
direct digitization until later. As microprocessors become more powerful
they will be capable of acting as slaves to an intelligent SUMEX program.
The AI program will direct the image processing functions of the micro so
that the data is processed in an intelligent way, but SUMEX will only see
the results of that processing, not the actual data. We will thus need to
keep track of developments in microcomputers so that we can develop this
kind of distributed system.

Privileged Communication 295 E. A. Feigenbaum
Ultrasonic Imaging Project Section 9.3.1

3. Recommendations

Since we are planning to develop a distributed system we would hope
to see these kind of systems being developed by the SUMEX resource.
Projects that would be of direct interest are networks (such as ETHERNET),
personal computer stations, graphics displays, etc.

E. A. Feigenbaum 296 Privileged Communication
Section 9.4 Pilot AIM Projects

9.4 Pilot AIM Projects

 

The following are descriptions of the informal pilot projects
currently using the AIM portion of the SUMEX-AIM resource or the Rutgers-
AIM resource pending funding, and full review and authorization.

Privileged Communication 297 E. A. Feigenbaum
Coagulation Expert Project Section 9.4.1
9.4.1 Coagulation Expert Project

Coagulation Expert Project

Donald Lindberg, M.D.
University of Missouri
Columbia, Missouri

I. SUMMARY OF RESEARCH PROGRAM
A. Project rationale

Preliminary experiment in attempting to form a clinical consultant
program based on a formal representation of medical knowledge of the blood
coagulation (or clotting) expert.

B. Medical relevance and collaboration

Experts in clotting are few and tend to be based at University
hospitals or large tertiary care facilities. It would be extremely helpful
if this knowledge could be made available to physicians via an automated
system.

Relevance of such a proposed system would be with respect to
diagnosis, management, and continuing medical education.

The team at the University of Missouri-Columbia consists of the
following individuals:

Lamont Gaston, M.D.

David Goldman

Lawrence C. Kingsland III
Donald A. B. Lindberg, M.D.
Haruki Ueno, Ph.D.

Anthony Vanker. Ph.D.

Dr. Gaston is a consulting hematologist, director of a coagulation
laboratory, and co-director of a blood-banking service.

Expertise in the field as well as clinical laboratory and patient
records are being provided by UMC to build and test the consultant. In the
future we plan to incorporate the views of external experts as well.

A formal research proposal to NIH is planned for fall, 1980, based on
the studies performed-on SUMEX.

C. Highlights
-Accomp1ishments

Use of UNITS/AGE: an initial model has been created on SUMEX.

E. A. Feigenbaum 298 Privileged Communication
Section 9.4.1 Coagulation Expert Project

Experimental use of EMYCIN: a feasibility test with a text book level
consultant model has been created on SUMEX.

Use of local LSI-11: in addition, the initial knowledge base has been
assembled into a simpler (but operational) system on a DEC LSI-11 using RT-
11 and BASIC.

We have selected a strategy for development. This is to begin with
the interpretation of clinical laboratory tests: first the full coagulation
screen (of 6 tests), then the partial coagulation screen (of 3 tests), and
finally the individual determinations. In all these cases, laboratory and
clinical features will be taken into account.

~Research in progress

Currently we are testing the initial models against actual clinical
records for 270 patients. This is partly as a validation of the work done,
and partly as a means to bring to our attention the unusual circumstances
and unforeseen problems which we know will be present. That is, we have
allowed for all feasible patterns of results, but (probably) have not yet
allowed for all the surrounding clinical circumstances. In any event, the
data gathering is almost complete and testing is about to begin.

D. List of relevant publications

None

E. Funding support

This preliminary research phase is being supported from two sources:

1. USPHS Grant No. T15 LM 07006, "Training Program in Medical
Information Science", Full funding is $162,410/year. About $25,000/year is
being devoted to this project.

2. USPHS Grant No. HS 02569, "Health Care Technology Center." Current
funding is $500,000/year. About $12,000/year is being devoted to this
project.

II. INTERACTIONS WITH THE SUMEX-AIM RESOURCE
A. Medical collaborations and program disseminations via SUMEX

Dr. Vanker will give an oral presentation of our work at the Spring
Meeting of Trainees and Directors, N.L.M. Training Programs, in May at
Columbus, Ohio. He also plans to demonstrate our AGE-implemented model
during the meeting. We have also given individual demonstrations of our
models to visiting scientists (including some from Japan) at UMC.

B. Sharing and interactions with other SUMEX-AIM projects

In February, 1980, David Goldman, a medical student at UMC and former
pre-doctoral fellow in the Information Science Group, spent a week at

Privileged Communication 299 E. A. Feigenbaum
Coagulation Expert Project Section 9.4.1

Stanford University becoming acquainted with the various artificial
.intelligence (AI) systems in development at SUMEX. In fact, with the help
of members of the SUMEX-AIM community, he was to implement a simple,
workable coagulation model in EMYCIN.

In March, 1980, Dr. Ueno attended a workshop on AGE at Stanford.
Through this workshop he was able to learn a great deal that was directly
applicable to our work. He also obtained a better understanding of the
UNITS package and how it might be used to interface with AGE. All of in the

Study group are planning to attend the AIM-tutorial at Stanford in August,
1980.

Since the AI systems in which we are interested are in some stage of
development on the SUMEX computer, and since partial documentation does
exist, we have been able to learn a great deal on our own by an
interactive, trial-and-error method.

Of course we have had many questions, and we have received prompt and
helpful information from various members of the SUMEX-AIM community via the
network electronic message system.

C. Critique of resource management

We have found the people at SUMEX to be uniformly helpful and more
than willing to aid us in our attempts to understand the various aspects of
AI in medicine. Both Mr. Goldman and Dr. Ueno were delighted with their
experiences at Stanford, and commented on the willingness of otherwise very
busy people to help them with their problems.

One of the drawbacks of SUMEX is that quite often the interaction is
slow. There have been days when we must wait up to several minutes between
exchanges between our terminal and SUMEX. This is apparently due to a high
average load on SUMEX at the time. We have had no other problems with the
resources at SUMEX and we feel the management has done a good job thus far.
III. RESEARCH PLANS

A. Plans for Summer, 1980

1. Continue assembling the knowledge data base, with emphasis on

documentation of the primary literature sources for the knowledge sources
(KS).

2. Continue learning the various aspects of UNITS/AGE and EMYCIN.

3. Continue comparing the two potentially complex models with the
inherently simpler microprocessor version.

4. Appoint a clinical test panel for consultation on development of
the next features.

5. Prepare the application to NIH.

E. A. Feigenbaum 300 Privileged Communication