Utah Workshop on the Applications of Intelligent and Adaptive Systems

Jerome Soller soller at asylum.cs.utah.edu
Fri Apr 16 15:47:30 EDT 1993


 	The University of Utah Cognitive Science Group's Industrial
Advisory Board presents the first Utah Workshop on the
"Applications of Intelligent and Adaptive Systems."  This will
be held at the Salt Lake City VA Medical Center's multipurpose
room, Building 8, on May 10th.  This event is free and open to the
public.  A listing of speakers, titles, abstracts, e-mail 
addresses, phone numbers, parking information, etc.., is provided 
at the end of this notice.  For listings of relevant publications, 
biographical sketches of the speakers, or a hard copy of a map of 
the VA campus and appropriate parking, contact Jerome Soller 
(soller at asylum.cs.utah.edu,582-1565, ext. 2469).  Hopefully, we
will have a videotape of this event available public domain
at a later point. 
	This workshop complements the yearly Psychology Department's
William F. Prokasy Lecture by Dr. Irving Biederman, Head of the
Cognitive and Behavioral Neuroscience Program at the University of
Southern California.  Dr. Biederman will speak on Tuesday, May 11, 
at 5:00 p.m. in BEH SCI 110 at the University of Utah.  The title is 
"Shape Recognition in Mind and Brain."  The official notice from
Kim Poulson is included after the workshop description.

	
Summary (abstracts and contact information are provided at the 
end of this notice):  
	Talk 1: 10:00-10:45
	The Utah State University Space Dynamics Laboratory, An Introduction

Presented by 	J Steven Hansen, Ph.D.
		Director, Instrument and Data Evaluation Center
			Space Dynamics Laboratory
		Associate Research Professor, 
			Department of Electrical and Computer Engineering
			Department of Physics
			Utah State University

	Talk 2: 10:45-11:30
	Active Noise and Vibration Control:
          Structures, Algorithms, and Applications

                      Scott C. Douglas, Ph.D.
            Department of Electrical Engineering
                     University of Utah

	Talk 3: 11:30 - 12:15
		A COMPUTERIZED DECISION SUPPORT SYSTEM FOR CRITICAL CARE:
MANAGEMENT OF MECHANICAL VENTILATION IN PATIENTS WITH ARDS
Thomas D. East, Ph.D.:  Director of informatics research in the pulmonary 
division at LDS Hospital(Intermountain Health Care).  Associate professor 
of anesthesiology, bioengineering and medical informatics at the University 
of Utah. 

	Lunch break: 12:15-1:45

	Talk 4: 1:45-2:30
		"Neural Networks for Classification, Signal Processing, and
Control in Patient Monitoring"
Dwayne Westenskow, Ph.D. Acting Director of Institute of 
Life Support in Space and Professor, U. of Utah Department of 
Anesthesiology (additional appointments in Bioengineering, Surgery, and
Medical Informatics).

	Talk 5: 2:30-3:15
		Towards a General Self-Organizing Learning Model
		Tony Martinez, Ph.D., Director of the Neural Networks and
Machine Intelligence Research Group, Brigham Young University
Assistant Professor, Department of Computer Science and Department of 
Electrical and Computer Engineering, Brigham Young University.
 
	Talk 6: 3:15-4:00
		Neil Cotter, Ph.D. Consulting Engineer, Geneva Steel
Research Assistant Professor, Electrical Engineering, University of Utah
"Neural Networks, Fuzzy Logic, and Triangulation in Process Control" 
(801-227-9865)
		
Sincerely,
	
	Dick Burgess, U. of Utah Department of Physiology and
		Program Director, U. of Utah Cog. Sci. Group
		(801-581-4072)
	Dale Sanders, Chairman of Industrial Advisory Board to U. of Utah
		Cognitive Science Group and Senior Technical Member,
		 TRW Corporation (dsanders at bmd.trw.com, 801-625-8343)
	Robert L. Angell, Principal, Applied Information and Management
		Systems and Small Business Representative to U. of Utah
		Cognitive Science Group (bangell at cs.utah.edu, 801-583-8544)
	Jerome B. Soller, VA GRECC and U. of Utah Department of Computer 
		Science  (soller at asylum.cs.utah.edu, 801-582-1565,
		ext. 2469)

Abstracts:
----------------------------------------------------------------------
             Active Noise and Vibration Control:
          Structures, Algorithms, and Applications

                      Scott C. Douglas, Ph.D.
            Department of Electrical Engineering
                     University of Utah

Active noise and vibration control is a method for cancelling 
unwanted sound or vibration by generating and introducing an 
equal-but-opposite acoustic signal into the noise environment. 
Current research and application developments in this field 
include active mufflers for automobiles, active quieting of 
machine noise on factory floors, active vibration mounts for 
airplane engines, active noise suppressors for air ducts, and 
active suppression of wall vibration for apartment dwellings. 
In this talk, I will provide an overview of the technology needed 
for active noise control systems.  In particular, I will discuss 
the digital signal processing hardware and multichannel adaptive 
control algorithms required for noise cancellation across large 
acoustic regions.  Results from an active noise suppression system 
for quieting an air conditioner compressor will be presented.  

------------------------------------------------------------------------
	Neil Cotter, Ph.D. Consulting Engineer, Geneva Steel
	Research Assistant Professor, E.E. Department, University of Utah
	"Neural Networks, Fuzzy Logic, and Triangulation in Process 
	Control" (801-227-9865)
	Many process control problems in industry require the approximation
	of nonlinear functions of several variables.  This presentation
	will include a tutorial showing that the following approximation
	methods are structurally similar to Fourier series: linear regression,
	neural networks, radial basis functions, and fuzzy logic.  The 
	speaker will contrast these methods with a practical interpolation
	method called triangulation.

--------------------------------------------------------------------------
	"Neural Networks for Classification, Signal Processing, and
Control in Patient Monitoring"
	
	Dwayne R. Westenskow, Ph.D.
	Department of Anesthesiology (801-581-6393, drw at cc.utah.edu)

	Artificial neural networks learn from examples to construct a 
function which maps input variables to a desired output.  The mapping
provides closed-loop control where the inputs are the variables to be
controlled, and the output is the drive signal for an actuator.  Neural
network control will be illustrated with the truck backing up example,
introducing the concept of supervised learning.  The neural network
mapping similarity provides for the classification of data, as will
be illustrated with an intelligent anasthesia alaram system.
Because the mapping function extrapolates between individual
data points, the neural network provides signal processing, i.e.,
filtering and noise and artifact rejection.  This will be illustrated
by showing the neural network processing of oscillometric blood 
pressure curves and cardiac output by thermal dilution.  In both of
these examples, the neural network provides a nonlinear mapping
which compensates for the overestimation at low values and 
underestimation at high values, which occurs with traditional
linear algorithms.  Success in processing Doppler signals for the
detection of air embolism illustrates how the mapping function
is constructed using training data rather than expert knowledge.
---------------------------------------------------------------------
Dr. Tony Martinez, Director of the Neural Networks and
Machine Intelligence Research Group, Brigham Young University
Assistant Professor, Department of Computer Science and Department of 
Electrical and Computer Engineering, Brigham Young University, Provo, UT
Towards a General Self-Organizing Learning Model
(martinez at cs.byu.edu, 801-378-6464)

A new class of connectionist architectures is presented called ASOCS
(Adaptive Self-Organizing Concurrent Systems).  ASOCS models support
efficient computation through self-organized learning and parallel
execution.  Learning is done through the incremental presentation of rules
and/or examples.  Data types include Boolean and multi-state variables;
recent models support analog variables.  The model incorporates rules into
an adaptive logic network in a parallel and self organizing fashion.  The
system itself resolves inconsistencies and generalizes as the rules are
presented.  After an introduction to the ASOCS paradigm, the talk
introduces current research thrusts which significantly increase the power
and applicability of ASOCS models.  Current application targets include
adaptive network routing, speech recognition, automated document updating,
and general classification problems.
------------------------------------------------------------------------

Author Information: Thomas D. East, Ph.D., Pulmonary Division, 
LDS Hospital, 8th Ave and C St., Salt Lake City, UT, 84143, 
801-321-3503 phone, 801-321-1671 fax, teast at fenta.med.utah.edu 
Audio Visual Media used: 35 mm slide projector and VHS video 
	projector.

A COMPUTERIZED DECISION SUPPORT SYSTEM FOR CRITICAL CARE:
MANAGEMENT OF MECHANICAL VENTILATION IN PATIENTS WITH ARDS
T.D. East, A.H. Morris, C.J. Wallace, A. T. Kinder, W.D. Littman*, 
	J.S. Gochberg*
Pulmonary Division, LDS Hospital,Salt Lake City, Utah 84143
* ACT/PC, 6501 Watts Road, Suite 115, Madison, WI 53719

Thomas D. East,Ph.D.:  Associate professor of anesthesiology, 
bioengineering and medical informatics at the University of Utah. 
Director of informatics research in the pulmonary division at LDS 
Hospital.  MEBE and Ph.D. in bioengineering from the U of U.  My 
research interests are in the applications of computers to critical 
care.  In particular much of my work has been in the area of rule 
based decision support systems and knowledge engineering.

	The care of critically ill patients is increasingly complex 
and clinicians frequently suffer from information overload.  It is 
difficult, if not impossible to assess all this information and 
generate a systematic and reasonable therapy plan.  Computerized 
decision support systems can assist the clinician with many of the 
tasks such as the iterative management of mechanical ventilation.  
This decision support not only standardizes care but may improve 
the quality of care by reducing mistakes.  This standardization 
of care also makes it possible to thoroughly characterize the 
current treatment process in order to compare it to a proposed 
new therapy as part of an ongoing continuous quality improvement 
(CQI) program.
	A computerized decision support system for the management 
of mechanical ventilation (respiratory evaluation, oxygenation, 
ventilation, weaning and extubation) in patients with adult 
respiratory distress syndrome has already been developed and 
clinically validated at the LDS Hospital  (1, 2) .  The protocol 
logic was developed using our existing consensus generating 
physician group and was implemented on the HELP system  (3) .  
The computerized decision support system was used for over 
35,000 hours in 111 Adult Respiratory Distress Syndrome 
patients and has controlled decision making 95% of the 24 hour day.  
The survival rate was 67%, higher than the expected 31-33% 
from historical data  (4, 5) , p < 0.05.  These results have 
demonstrated that computerized decision support for critical 
care is feasible.
	We are in the process of conducting a prospective 
randomized clinical trial to test efficacy of 
computerized protocols in 400 patients with ARDS at two different 
clinical sites; KDMC a county hospital in the Watts district of 
Los Angeles, CA and Hermann Hospital, a private hospital in 
Houston affiliated with University of Texas Medical School 
(H0: There is no difference in efficacy between protocol and 
non-protocol controlled critical care).  The knowledge base 
(set of protocol logic rules) was transferred from the HELP 
system at LDS Hospital to a PC based ICU computer system known 
as ARGUS Windows (ACT/PC, Madison , WI).  ARGUS Windows runs 
under QNX V4.1 and QNX windows V2.03.  The rules were implemented 
using a rule based decision support engine designed by ACT/PC.   
The engine is a finite state automata written in C for QNX and 
QNX Windows.  The ARGUS Windows system has been installed at 
all 12 beds of the surgical ICU at KDMC.  This system is now 
in routine use for respiratory care charting and the decision 
support system has been used to successfully care for 10 
ARDS patients in a pilot study of feasibility.  In the 
randomized trial, will define efficacy using a hierarchical 
four level approach; Efficacy F a)Survival, b) Length of 
ICU Stay, c) Morbidity, d) Incidence and severity of barotrauma.  
Generalizablity of the computerized decision support system will be 
determined by examining; 1) Percent of total time in the trial 
during which protocols controlled patient care. 2) Number of 
protocol instructions which were not followed. 3) Number of 
objections to protocol logic which, based on medical evidence, 
forced a change in the logic.  To our knowledge this is the first 
prospective randomized clinical trial designed to test the 
impact of computerized critical care decision support on 
patient outcome.
References:
1.	East T, et al.  Int J Clin Monit Comput 1992;8:263-269.
2.	Morris AH, et al.  Am Rev Respir Dis 1992;145(4):A184.
3.	Pryor TA, et al. The HELP system development tools. 
    In: Implementing health care information 
    systems.  New York: Springer-Verlag, 1989: 365-383. 
4.	Zapol WM, et al. The adult respiratory distress syndrome 
    at Massachusetts General Hospital, Etiology progession and 
    survival rates,1978-1988. In: Zapol WM, Lemaire F, ed.  
    Adult Respiratory Distress Syndrome.  New York: Marcel Dekker, 
    Inc, 1991: 367-380. 
5.	Artigas A, et al. Clinical presentation, prognostic 
    factors, and outcome of ARDS in the European Collaborative 
    Study (1985-1987).In:Same book as ref 4, 1991: 37-63. 
Acknowledgements:  This work was supported by NHLBI grant #HL36787, 
AHCPR grant HS06594, Siemens Ventilators, ACT/PC, the Respiratory 
Distress Syndrome Foundation and the Deseret Foundation (LDS Hospital).

My biographical sketch:

Thomas D. East,Ph.D.:  Associate professor of anesthesiology, 
bioengineering and medical informatics at the University of Utah. 
Director of informatics research in the pulmonary division at LDS 
Hospital.  MEBE and Ph.D. in bioengineering from the U of U.  My research 
interests are in the applications of computers to critical care.  
In particular much of my work has been in the are of rule based 
decision support systems and knowledge engineering.
------------------------------------------------------------------------

The Utah State University Space Dynamics Laboratory, An Introduction

Presented by 	Dr. J Steven Hansen (801-750-4850, jsh at sdl.usu.edu)
		Director, Instrument and Data Evaluation Center
			Space Dynamics Laboratory
		Associate Research Professor, 
			Department of Electrical and Computer Engineering
			Department of Physics
			Utah State University, Logan, Utah

Abstract

	This talk will provide description of the Space Dynamics Laboratory
and its capabilities, particularly in the areas of data, signal, and image
processing.  The presentation will include a brief history of SDL over the 
past 30 years, an overview of the past and current projects at SDL and the 
Instrument and Data Evaluation Center (IDEC), a look at the current computing
and image/signal processing capabilities IDEC. A look at the future and 
particular area of possible interest to the community will be addressed.

------------------------------------------------------------------------
	Lunch: A group of us will go to lunch at a our own expense.
If you wish to join that group, contact Jerome Soller to have a spot reserved.

	Directions to the VA Hospital:
For directions, call 582-1565, and ask for learning resources.
The address is 500 Foothill Boulevard

	Parking:  The VA has limited parking for visitors behind
building 8.  If you come in the main entrance, make a right past building
1 and 14, continue through one of the staff parking lots.  Make 
a right before reaching the staff parking lot between building
5 and 8.  You should pass between buildings 8 and 9, and park behind
building eight (not nine).
Rough Map:  Dash lines represent a desired path.

Main Entrance:--   1
		|
		|  14
		|		5
Don't Park	|	    	Don't Park Here    
Here		|   ----------|
		|---|Don't    |	
		     Park     |
		     Here     |
			   9  |    8       6
			      |___	   Don't Park 
		     Don't Park   Park 		Here
			Here      Here
			


It is advisable to have a map of the VA campus mailed to
you, and that map and postage will be free of charge if Jerome
Soller is notified within the next two weeks.  If parking is tight, 
you may need to park nearby in Research Park (leave ample time for that).  
If you are at the University of Utah, you can catch a VA shuttle from the 
medical school or research park directly to the VA hospital.  University 
shuttles run to university parking, which is across from the VA hospital,
and to the LDS Insitute, which is also across the VA hospital.

	Registration: No registration is required.  However, it would
be greatly appreciated if you would rsvp by the end of April
so we can estimate the attendance and set up the room accordingly.
The room can comfortably seat 80 people with good workspace.

	Afterwards: We hope that speakers will provide some form
of handouts.  We hope to have a video tape record of this
workshop, which will be public domain.

***************************************************************************


More information about the Connectionists mailing list