Joshua - Stanford Universityfb827dx0405/fb827dx0405.pdf · Kraut, Dr. Stephan Freer and other...

«

-*

TIhe Stanford University Medical Experimental

Computer (SUMEX) was established in January,1974, to constitute the first national sharedcomputing resource for medical research. Aninnovative effort to help biomedical scientists meettoday's research requirements and to explorecomputer applications in many health fields rangingfrom basic research to bedside care, SUMEX isdirected by Professor Joshua Lederberg, Chairman ofStanford's Department of Genetics. The project hasbeen funded by a grant from the Division of ResearchResources of the National Institutes of Health(Biotechnology Resources Program) for an initialterm that expiresin July, 1978.

At present, SUMEX consists of a powerfultime-shared dual processor DEC-10 computer system.It is available to approved users throughout theUnited States over computer communicationsnetworks. The project's goals for its present 5-yearterm are: 1) the encouragement of applications ofartificial intelligence in medicine (AIM), and 2) themanagerial, administrative and technicaldemonstration of a nationally-shared technologicalresource for health research.

Such a resource offers scientists both a significanteconomic advantage in sharing expensiveinstrumentation and a greater opportunity to shareideas about their research. This is especially timely incomputer science, a field whose intellectual andtechnological complexity tends to nurture relativelyisolated research groups. Each group may then tendto pursue its own line of investigation with limitedconvergence on working programs available fromothers. In this respect, computer applications havedemonstrated less mutual incremental progress fromdiverse sources than is typical of other sciences. TheSUMEX-AIM project seeks to reduce thesebarriers toscientific cooperation in the field of artificialintelligenceapplied to health research.

Each authorized project in the SUMEX-AIMcommunity is concerned in some way with theapplication of these principles to medical and healthresearch problems. This type of "intelligent"assistance by computer program is perhaps bestillustrated by the following brief descriptions of aselected sample of SUMEX-AIM projects.

ARTIFICIAL INTELLIGENCE

The term "artificial intelligence" (Al) refers toresearch efforts aimed at studying and mechanizinginformation-processing tasks that have required theapplication of some degree of human intelligence.Controversial speculations on how far this eventuallymay lead only distract from pragmatically usefulapplications of currently feasible art. The currentemphasis in the field is to understand the underlyingprinciples of a) efficient acquisition and utilizationofmaterial knowledge, and b) the programmedrepresentation of conceptual abstractions inreasoning, deductive, and problem-solving activities.At present, these are far more specialized andinflexible than human intellectual functions;however, in special domains they may be ofcomparable or greaterpower, e.g., in the solution offormal problems in organic chemistry or in theintegral calculus.

DENDRAL

The DENDRAL project at Stanford, under thedirection of Professor Lederberg, Genetics; ProfessorEdward Feigenbaum, Computer

Science;

andProfessor Carl Djerassi, Chemistry, is aimed atassisting the biochemist in interpreting molecularstructures from spectroscopic, physical and chemicalinformation. In cases where the characteristicspectraof a compound are not catalogued in libraries, theDENDRAL programs carry out the rather laboriousprocesses a chemist must go through to interpret thespectrum from "first principles." One of theDENDRAL programs, CONGEN (for CONstrainedstructure GENeration), is an interactive programdesigned to assist the chemist in the enumeration ofstructural isomers, based on inferences aboutstructural features of an unknown compound. Theseinferences, whether obtained from physical, chemicalor spectroscopic data, are supplied to CONGEN asstructural fragments and related information, using astandard computer terminal connected toSUMEX-AIM. The program uses atoms andsuperatoms (non-overlapping structural fragmentsknown to be present in the molecule) to constructstructures; the procedure is restricted by a variety ofconstraints on desired and undesired substructuresand ring systems. There is no direct algorithmic pathavailable to determine such a molecular structurefrom the spectral data—only the inferential process ofhypothesis generation and testing within the domainof reasonable solutions defined by a knowledge oforganic and physical chemistry.

"

AI systems are characterized by complexcomputational processes that are primarilynon-numeric, e.g., graph-searching and symbolicpattern analysis. They involve procedures whoseexecution is controlled by diverse types and forms ofknowledge about a given task domain, such asmodels, fragments of "advice," and systems ofconstraints or heuristic rules. Unlike conventionalalgorithms commonly based on a well-tailoredmethod for a given task, AI procedures typically use amultiplicity of methods in a highly conditionalmanner-depending on the specific data in the taskand a variety of sources of relevant information. Thetangible objective of this approach is the practicaldevelopment of computer programs which, usingformal and informal knowledge together withmechanized hypothesis formation andproblem-solving procedures, will offer more generaland effective consultative tools for the clinician andmedical scientist. Contexts in which experimentaldata already are acquired by machine may offer evenricher opportunities.

This process, as implemented in the computer, is asimplified example of the cycle of inductivehypothesis-deductiveverification that is often taughtas a model of the scientific method. (Whether this is afaithful description of contemporary science is

*

(called the ONET) for computer diagnosis andtreatment of glaucoma. The computer system, whichincludes an elaborate pathophysiologic model of thedisease, is being tested through the SUMEX-AIMnetwork at five eye centers: Mount Sinai Hospital andMedical Center, New York; Washington University,St. Louis; The Johns Hopkins University, Baltimore;the University of Illinois at Chicago Circle; and theUniversity of Miami. Glaucoma, in one form oranother, affects 2% ofall peopleover40 years of age.It is a disease in which increased pressure within theeye may lead to irreparableoptic nerve damage andblindness. The computer-based program has greatpotential for assisting clinicians and researchers inunderstanding the disease, diagnosing it moreaccurately and improving its treatment.

arguable, and how it may be implemented in thehuman brain is unknown. Regardless, these are usefulleads rather than absolute preconditions for thepragmatic improvement of mechanized intelligencefor more efficient problem-solving.) The elaborationof these approaches with existing hardware andsoftware technologies is the mostpromising approachto enhancing the application of computers to thevaguely structured problems that dominate our taskdomains.

A new pilot project, MOLGEN, has beenmotivated by the success of the DENDRAL effort.MOLGEN uses similar paradigms in an effort tomechanize experiment-planning in molecular genetics,particularly work on DNA structure and inter-speciestransfer being conducted in Professor Lederberg'slaboratory. Whereas the DENDRAL goal was ahypothesis(e.g., a chemical structure) to explaina setof experimental data, MOLGEN begins with astipulated DNA structure and seeks suggestedexperiment plans that could either falsify or validatethe asserted structure. At present, this entails asubstantial effort in representing existing knowledgeof experimental techniques(e.g.,enzyme specificities,electron-microscopy, electrophoresis) and thephysical biochemistry of DNA.

In another project, Professor Charles Schmidt, asocial psychologist, is developing a theory of howpeople arrive at interpretations of the social actionsof others in collaboration with Professor N. S.Sridharan, a computer scientist. The theory will betested in situations such as the psychiatric interviewand the legal trial. The computer system whichcurrently represents the theory is called "Believer." Itincludes a large body of statements about people'smotivationsand actions.

The Rutgers project includes, in addition, severalfundamental studies in artificial intelligence andsystem design. These provide much of the supportneeded for the developmentof complex systems suchas the glaucoma consultation and the "Believer"programs.THE RUTGERS PROJECT

COMPUTERS IN BIOMEDICINEProfessor Saul Amarel, a Rutgers University

computer scientist, directs several research effortsdesigned to introduce advanced methods in computerscience-particularly in artificial intelligence andinteractive data base systems-into specific areas ofbiomedical research.

SIMULATION AND EVALUATIONOF CHEMICAL SYNTHESIS

The development of new drugs and the study ofhow drug structure is related to biological activitydepends on the chemist's ability to synthesize newmolecules and modify existing structures, e.g.,incorporating isotopic labels into biomolecularsubstrates. The Simulation and Evaluation ofChemical Synthesis (SECS) project, directed by Dr.Todd Wipke, Associate Professor of Chemistry at the

For example, a group led by Professor CasimirKulikowski is developing computer-basedconsultation systems for diseases of the eye incollaboration with Dr. Aran Safir, an ophthalmologistfrom the Mount Sinai School of Medicine. Animportant development in this area is theestablishment of a national network of collaborators

*

University of California, Santa Cruz, is aimed atassisting the synthetic chemist in designingstereospecific syntheses of complex bio-organicmolecules.

MYCINComputer-based Consultation

in Clinical Therapeutics

The molecule to be synthesized is presented toSECS using interactive computer graphics. Theprogram studies the chemical graph and alsothree-dimensional and electronic models of themolecule which it knows how to construct; then,using fundamental chemical principles, and variousheuristics, it works backwards from the target topredict possible precursors which are one syntheticstep away from the target. The chemist selects theprecursors to be considered by the program forfurther analysis. Thus, SECS acts as a consultant,working with the chemist to form achemist-computer team. The chemist helps guide thesearch and decides when to stop the analysis.Knowledge about chemical transformations isexpressed directly by chemists in ALCHEM, anEnglish-like language interpreted by SECS. Goals forfurther developmentof the project include generationof constraining strategies based on symmetry, stericand electronic considerations, and expansion of thechemical transform data base.

Dr. Stanley Cohen, Professor and Head of theDivision of Clinical Pharmacology at Stanford, directsthis research in collaborationwith Dr. Stanton Axlineand with computer scientists interested in artificialintelligence and medical computing. The MYCINsystem models the decision processes of medicalexperts, utilizing both clinical data and thejudgmental knowledge of experts to providephysician nonspccialists with consultative adviceregarding clinical therapeutics. Although initialresearch concerns the use of antimicrobial agents inthe treatment of bacteremias, the system is beingexpanded to deal with the treatment of otherinfections.

J

The primary component of the system is theConsultation program which uses the physician'sresponse to computer-generated questions about apatient to make deductions about the case. It thenadvises the physician on the infectious diseasediagnosis and the recommended treatment for thepatient. The utility and flexibilityof thisprogram areincreased by three adjunct programs: 1) aQuestion-Answering program which answersquestions about the system's knowledge base andabout a specific consultation, 2) an Explanationprogram which justifies the consultative advice andexplains the system's deduction process, and 3) aKnowledge Acquisition program which extends theknowledge base of the system through dialoguewithan expert.

In addition to its ongoing development on theSUMEX-TYMNET system, an experimentalversion ofSECS is available over TELENET from First DataCorporation in Waltham, Massachusetts. SECS alsoruns on a Univac system at the University ofStrasbourg, France, and on PDP-1 0's at theUniversities of Darmstadt and Heidelberg, Germany.Feedback from this outside use of SECS spotlightsareas for needed workand provides positive evidenceof the usefulness of SECS as a tool in syntheticdesign.

Goals for further development of the systeminclude implementation and evaluation of the systemin the clinical setting at the Stanford University andPalo Alto Veterans Administration Hospitals.

I1

ACTA Model of Human Cognition

The ACT project is directed by Dr. JohnAnderson, Associate Professor of Psychology at YaleUniversity. The ACT program provides a uniform setof theoretical mechanisms to model such aspects ofhuman cognition as memory, inferential processes,language processing, and problem-solving. Theknowledge base consists of two components. Thepropositional component is provided by anassociative network encoding a set of known factswhich provide the system's semantic memory. Theprocedural component consists of a set ofproductions which operate on the associativenetwork. The production system used is considerablydifferent than those in other currently availablesystems, e.g., Newell's PSG, and allows the system tooperate on an associative network and to moreaccurately model certain aspects of human cognition.

ARTIFICIAL INTELLIGENCEMETHODOLOGY APPLIED TO

PROTEIN CRYSTALLOGRAPHY

Members of the artificial intelligence project atStanford also are collaboratingwith Professor JosephKraut, Dr. Stephan Freer and other proteincrystallographcrs at the University of California, SanDiego. They arc using the SUMEX-AIM facility as thecentral repository for programs, data and otherinformation of common interest. The generalobjectives of the project are: 1) to identify criticaltasks in protein structure elucidation which maybenefit by the application of AT problem-solvingtechniques, and 2) to design and implementprogramsto perform those tasks.

Two principal task areas have been identifiedwhere collaboration is of practical and theoreticalinterest to both protein crystallographcrs andcomputer scientists working inAI: 1) interpreting athree-dimensional electron density map, and 2)determining a plausible structure in the absence ofphase information normally inferred fromexperimental isomorphous replacement data.

INTERNIST

The INTERNIST project, under the direction ofDr. Harry Pople and Dr. Jack Myers at the Universityof Pittsburgh, is a large-scale, computerized medicaldiagnostic system utilizing the methods andstructures of artificial intelligence. Unlike mostcomputer diagnostic programs, which are oriented todifferential diagnosis in a rather limited area, theINTERNIST system deals with the general problemof diagnosis in internal medicine and currentlyaccesses a medical data base encompassingapproximately 50% of the major diseases in internalmedicine.

MISL

The Medical Information Systems Laboratory atthe University of Illinois, Chicago Circle Campus, hasbeen established under the direction of Dr. BruceMcCormick of the Department of InformationEngineering, in collaboration with Dr. MortonGoldbergof the Department of Ophthalmologyat theUniversity of Illinois Medical Center.

The foremost goal of the resource is theexploration of artificial intelligence techniques inautomated clinical decision-making inophthalmology. Investigations into the constructionof a data base in ophthalmology,and into distributeddata base design, are ancillary goals. Incorporatingreliable clinical information into the ophthalmologydata base is a critical prerequisite to adequateclinicaldecision support. Core research concerns theexploration of inferential relationships betweenanalytic data and the natural history of selected eyediseases, both in treated and untreated form.

MISL utilizes the computer facilities of theUniversity of Illinois and the SUMEX-AIM network,providing the administrative structure for assemblingthe expertise of the collaborating departments.Serving as a bridge between diverse academic worlds,MISL promotes close involvement betweenengineeringand medical faculty. The Illinois Eye andEar Infirmary at the Medical Center, with athroughput of 50,000 patientsper year, provides anideal setting for the direct application of computertechnology to real problems in clinical medicine.

SUMEX-AIM Management

A significant part of the SUMEX-AIM experimenthas been the developmentof a managementstructureto maximize the utility of the computer capabilityfor a national community.

Users of the SUMEX facility are divided foradministrative purposes into two groups: 1) local, atStanford University School of Medicine, and 2)national, elsewhere in the United States. As PrincipalInvestigator for the SUMEX grant, Dr. Lederbergreviews Stanford medical school projects with theassistance of a local advisory committee. Nationalusers may gain access to the facility through anadvisory panel for a national program in ArtificialIntelligence in Medicine (AIM). The AIM AdvisoryGroup consists of members-at-large of the AI andmedical communities, facility users and the PrincipalInvestigator of SUMEX as an ex-officio member. Arepresentative of the National Institutes ofHealth-BiotechnologyResources Program (NIH-BRP)serves as Executive Secretary.

The SUMEX-AIM computing resource is allocatedinitially to qualified users without fee. This, ofcourse, entails a careful review of the merits andpriorities of proposed applications. At the directionof the Advisory Group, expenses related tocommunications and transportation to allow specificusers to visit the facility also may be covered.

SUMEX-AIM is aware of the necessity of makingthe central facility available for trial use by potentialusers and collaborators. A GUEST mechanism hasbeen established for those who have an indicatedrequirement for brief access to certain programs.Those who have been given an appropriatetelephonenumber and login procedure can dial up SUMEX-AIMto exercise these programs on a trial basis. A specificobjective of many user projects is the demonstrationof their programs for thebenefit of a highly dispersednational community.

USER QUALIFICATIONSThe SUMEX-AIM facility is a community effort,

not merely a machine service. Applications formembership are judged on the basis of the followingcriteria:

1) The scientific interest and merit of theproposed research and its relevance to thehealth research missions of the NIH.

2) The congruence of research needs and goals tothe Al functions ofSUMEX-AIM as opposed toother computing alternatives.

3) The user's prospective contributionsand role inthe community, with respect to computerscience, e.g., developing and sharing newsystems or applications programs, sharing useof special hardware, etc.

4) The user's potential for substantive scientificcooperation with the community, e.g., to shareexpert knowledge in relevant scientificspecialties.

5) The quantitative demandsfor specific elementsof the SUMEX-AIM resource, taking accountof both mean and ceiling requirements.

FACILITY INFORMATIONThe computer facility consists of dual DEC Model

KI-10 CPU's running under a locally-developeddualprocessor TENEX operating system. It has 256Kwords (36-bit) of high-speed memory, 1.6M words ofswapping storage, 70M words of disk storage, two9-track 800 bpi industry-compatible tape units, a dualDEC-tape unit, a line printer, andcommunications network interfaces providing userterminal access. SUMEX may be accessed by localtelephone lines, through the TYMNET and as a hostover the ARPANET communicationsnetwork.

Program (software) support will evolve from thebasic system as dictated by the research goals andneeds of the user. Initially, availableprograms includea variety of TENEX user, utility and text editorprograms. Major user languages include INTERLISP,SNOBOL, SAIL, FORTRAN-10,

BLISS-10,

BASIC,Macro-10, OMNIGRAPH and MLAB.

Revised May 1976

POTENTIAL USERS

Potential users seeking further informationare invited to write:

Elliott Levinthal, Ph.D.AIM User LiaisonSUMEX-AIM Computer Projectc/o Department of Genetics, 5047Stanford University Medical CenterStanford, California 94305Telephone: (415)497-5813

Procedures for access to SUMEX-AIMare governed by the:

Biotechnology Resources ProgramDivision of Research ResourcesNational Institutes of HealthBuilding 31, Room 5819Bethesda, Maryland 20014

jiillii, I 1iiI j

ll llii'iiiii, xi!!I lull! 1, illII * 11II I

ii2ii iiiliiiliilil

I!

■im,

IH. Jill1tflia illifwill" f|j!

1.

illill, ii

J !!!!!! I 1j i mi

[ill ,i111, Jii 'mil .

1iiii

I!!!iii!li!!il!l!!iiii

»

TIhe Stanford University Medical Experi-

mental Computer (SUMEX) was established inJanuary, 1974, to provide the first shared na-tional computing facility for medical research.Directed by Dr. Joshua Lederberg, Professor andChairman of the Department of Genetics,SUMEX is an innovative effort to help biomedi-cal scientists meet today's research requirementsand to explore computer applications in manyhealth fields ranging from basic research tobedside care. The project is funded by a grantfrom the Division of Research Resources of theNational Institutes of Health (BiotechnologyResources Branch) for an initial term that ex-pires in July, 1978.

At present, SUMEX consists of a powerfulPDP-10 computer available to approved usersthroughout the United States over a computercommunications network on a time-shared basis.The project's goals over the next 5 years are: 1)the encouragement of applications of artificialintelligence in medicine (AIM), and 2) the man-agerial, administrative and technical demonstra-tion of a nationally-shared technological re-source for health research.

Such a resource offers scientists both a signifi-cant economic advantage in sharing expensiveequipment and a greater opportunity to shareideas about their research. This is especially truein computer science, a field whose intellectualand technological complexity has made it diffi-cult to avert the development of relatively iso-lated research groups. Each group may then tendto pursue its line of investigation with limitedconvergence on working programs available fromothers. The SUMEX-AIM project seeks to lowerthese barriers to scientific cooperation in thefield of artificial intelligence applied to healthresearch.

ARTIFICIAL INTELLIGENCE DENDRAL

The term "artificial intelligence" (AI) refersto research efforts aimed at studying and mecha-nizing information processing tasks that general-ly have been considered to require human intel-ligence. The current emphasis in the field is tounderstand the underlying principles in efficientacquisition and utilization of material know-ledge and representation of conceptual abstrac-tions in reasoning, deductive and problem-solving activities. AI systems are characterizedby complex computational processes that areprimarily non-numeric, e.g., graph-searching andsymbolic pattern analysis. They involve proce-dures whose execution is controlled by differenttypes and forms of knowledge about a given taskdomain, such as models, fragments of "advice"and systems of constraints or heuristic rules.Unlike conventional algorithms commonly basedon a well-tailored method for a given task, AIprocedures typically use a multiplicity of meth-ods in a highly conditional manner— dependingonthe specific data in the task and a variety ofsources of relevant information. The tangibleobjective of this approach is the production ofcomputer programs which, using formal andinformal knowledge together with mechanizedhypothesis formation and problem-solving pro-cedures, will be more general and effective con-sultative tools for the clinician and medicalscientist.

The DENDRAL project at Stanford, underthe direction of Dr. Lederberg, ProfessorEdward Feigenbaum, Computer Science, andProfessor Carl Djerassi, Chemistry, is aimed atassisting the biochemist in interpreting molecu-lar structures from mass spectral and otherchemical information. In cases where the charac-teristic spectrum of a compound is not cata-logued in a library, the DENDRAL programscarry out the rather laborious processes achemist must go through to interpret the spec-trum from "first principles." By symbolicallygenerating "reasonable" candidate structuresfrom hints within the spectrum and a knowledgeof organic chemistry and mass spectrometry, theprogram infers the unknown structure to be theone which best explains the observed spectrum.There is no direct algorithmic path available todetermine such a molecular structure from thespectral data—only the inferential process ofhypothesis generation and testing within thedomain of reasonable solutions defined by aknowledge of organic and physical chemistry.

This process, as implemented in the com-puter, is a simplified example of the cycle ofinductive hypothesis—deductiveverification thatis often taught as a model of the scientificmethod (Whether this is a faithful descriptionof contemporary science is arguable, and how itmay be implemented in the human brain isunknown. Regardless, these are useful leadsrather than absolute preconditions for the prag-matic improvement of mechanized intelligencefor more efficient problem-solving.). The elabo-ration of these approaches with existing hard-ware and software technologies is the mostpromising approach to enhancing computerapplication to the vaguely structured problemsthat dominate our task domains.

Each authorized project in the SUMEX-AIMcommunity is concerned in some way with theapplication of these principles to medical prob-lems. This type of "intelligent" assistance bycomputer program is perhaps best illustrated bythe following brief descriptions of someSUMEX-AIM projects.

*

THE RUTGERS PROJECTCOMPUTERS IN BIOMEDICINE

Professor Saul Amarel, a Rutgers Universitycomputer scientist, directs several researchefforts designed to introduce advanced methodsin computer science—particularly in artificial in-telligence and interactive data base systems—intospecific areas of biomedical research.

For example, a group of computer scientistsled by Professor Casimir Kulikowski is develop-ing computer-based consultation systems for di-seases of the eye in collaboration with Dr. AranSafir, an ophthalmologist from the Mount SinaiSchool of Medicine. An important developmentin this area is the establishment of a nationalnetwork of collaborators for computer diagnosisand treatment of glaucoma. The computersystem, which includes an elaborate pathophysi-ologic model of the disease, is being testedthrough the SUMEX-AIM network at three eyecenters: Mount Sinai Hospital and Medical Cen-ter, New York; Washington University, St.Louis; and The Johns Hopkins University, Balti-more. Glaucoma, in one form or another, affects2% of all people over 40 years of age. It is adisease in which increased pressure within theeye may lead to irreparable optic nerve damageand blindness. The computer-based program hasgreat potential for assisting clinicians and re-searchers in understanding the disease, diag-nosing it more accurately and improving itstreatment.

In another project, Professor Charles Schmidt,a social psychologist, is developing a theory ofhow people arrive at interpretation of the socialactions of others. The theory will be tested insituations such as the psychiatric interview andthe legal trial. The computer system whichcurrently represents the theory is called "Believ-er." It includes a large body of statements aboutpeople's motivations and actions. The SUMEX-AIM environment provides an excellent mediumfor collaboration between Dr. Schmidt's groupand other researchers around the country in the

development and testing of this computer-basedtheory.

The Rutgers project includes, in addition,several fundamental studies in artificial intelli-gence and system design. These provide much ofthe support needed for the development ofcomplex systems such as the glaucoma consulta-tion and the "Believer" programs.



Dr. Stanley Cohen, Associate Professor andHead of the Division of Clinical Pharmacology atStanford, directs this research in collaborationwith Dr. Stanton Axline and with computerscientists interested in artificial intelligence andmedical computing. An evolving computer pro-gram developed to assist physician nonspecialistsin the selection of therapy for patients withbacterial infections, MYCIN attempts to modelthe decision processes of medical experts. Itconsists of three closely integrated components:the Consultation System asks questions, makesconclusions and gives advice; the ExplanationSystem answers questions from the user tojustify the program's advice and explain itsmethods; and the Rule-Acquisition System per-mits the user to teach the system new decisionrules or to alter pre-existing rules judged to beinadequate or incorrect. Goals for further devel-opment of the system include expansion of theconsultation program to deal with infections

Iother than bacteremias and implementation andevaluation of the system in the clinical setting atStanford University Hospital.

COMPUTING APPLIED TOPROTEIN CRYSTALLOGRAPHY

Members of the artificial intelligence projectat Stanford also are collaborating with ProfessorJoseph Kraut and Dr. Stephan Freer, protein

crystal lographers at the University of California,San Diego. They are using the SUMEX-AIMfacility as the central repository for programs,data and other information of common interest.The general objective of the project is to applyproblem-solving techniques, which have emergedfrom artificial intelligence research, to the well-known "phase problem" of x-ray crystallo-graphy in order to determine the three-dimen-sional structures of proteins. The work is in-tended to be of both practical and theoreticalvalue to computer science (particularly artificialintelligence research) and protein crystallo-graphy.


A significant part of the SUMEX-AIM experi-ment is the development of a managementstruc-ture to maximize the utility of the computercapability for a national community.

Users of the SUMEX facility are divided foradministrative purposes into two groups: 1)those at Stanford University School of Medicine,and 2) those elsewhere in the United States. Thefacility resources (computing capacity and con-sulting support) are allocated in equal portionsto the two groups. As Principal Investigator forthe SUMEX grant, Dr. Lederberg reviews Stan-ford medical school projects with the assistanceof a local advisory committee. National usersmay gain access to the facility resources throughan advisory panel for a national program inartificial intelligence in medicine (AIM). TheAIM Advisory Group consists of members-at-large of the AI and medical communities, facili-ty users and the Principal Investigator ofSUMEX as an ex-officio member. A representa-tive of the National Institutes of Health-Biotech-nology Resources Branch (NIH-BRB) serves asExecutive Secretary.

DIALOG

The DlAgnostic LOGic project, under thedirection of Dr. Harry Pople and Dr. Jack Myersat the University of Pittsburgh, is a large-scale,computerized medical diagnostic system uti-lizing the methods and structures of artificialintelligence. Unlike most computer diagnosticprograms, which are oriented to differentialdiagnosis in a rather limited area, the DIALOGsystem deals with the general problem of diag-nosis in internal medicine and currently accessesa medical data base encompassing approximately50% of the major diseases in internal medicine.

The SUMEX-AIM computing resource is ini-tially allocated to qualified users without fee.This, of course, entails a careful review of themerits and priorities of proposed applications.At the direction of the Advisory Group, ex-penses related to communcations and transpor-tation to allow specific users to visit the facilitymay be covered as well.

MISLThe Medical Information Systems Laboratory

at the University of Illinois at Chicago Circle hasbeen established under the direction of Dr.Bruce McCormick, Information Engineering, incollaboration with Dr. Morton Goldberg, anophthalmologist at the U of I medical school.The project explores inferential relationshipsbetween analytic data and the natural history ofselected eye diseases both in treated and un-treated forms. SUMEX-AIM will be utilized tobuild a data base to be used as a test bed for thedevelopment of clinical decision support algo-

SUMEX-AIM is aware of the necessity ofmaking the facility available for trial use bypotential users and collaborators. A GUESTmechanism has been established for those whohave an indicated requirement for brief access tocertain programs. Those who have been given anappropriate telephone number and login proce-dure can dial up SUMEX-AIM to exercise theseprograms on a trial basis.

rithms.

e

USER QUALIFICATIONSApplications for use of the SUMEX-AIM facil-

ity are judged on the basis of:I)The scientific interest and merit of the

proposed research.2) The relevance of the research to the artifici-

al intelligence approach of SUMEX-AIM asopposed to other computing alternatives.

3) The user's prospective contributions androle in the community, e.g., developing andsharing new systems or application pro-grams, sharing use of special hardware, etc.

4) The user's capability and intentions of ope-rating in a community-effective style formutual advantage. Besides the program-ming innovations that some participantsmay contribute, all are expected to furnishexpert knowledge and advice about theexisting art in their fields of interest.

5) The quantitative allocation of specific ele-ments of the SUMEX-AIM resource basedon a concept of mean and ceiling plannedexpectations.

FACILITY INFORMATION

The computer facility, consisting of a DECModel KI-10 CPU running under the TENEXoperating system, has 256K words (36-bit) ofhigh-speed memory, 1 .6M words of swappingstorage, 70M words of disk storage, two 9-track800 bpi industry-compatible tape units, a dualDEC-tape unit, a line printer, and communica-tions-network interfaces providing user terminalaccess. SUMEX is available through TYMNETand as a host over the ARPANET communica-tions network.

Program (software) support will evolve fromthe basic system as dictated by the researchgoals and needs of the user. Initially, availableprograms include a variety of TENEX user,utility and text editor programs. Major userlanguages include INTERLISP, SNOBOL,SAIL,FORTRAN-10, BLISS-10, BASIC, Macro-10,OMNIGRAPHand MLAB.

Revised May 1975

POTENTIAL USERS

For further information, write:

Elliott Levinthal, Ph.D.AIM User LiaisonSUMEX-AIM Computer Projectc/o Department of Genetics, 5047Stanford University Medical CenterStanford, California 94305

Procedures for access to SUMEX-AIM aregoverned by the:

Biotechnology Resources BranchDivision of Research ResourcesNational Institutes of HealthBuilding 31, Room 58199000 Rockville PikeBethesda, Maryland 20014

"

TIhe Stanford University Medical Experi-

mental Computer (SUMEX) was established inJanuary, 1974, to provide the first shared na-tional computing facility for medical research.Directed by Dr. Joshua Lederberg, Professor andChairman of the Department of Genetics,SUMEX is an innovative effort to help biomedi-cal scientists meet today's research requirementsand to explore computer applications in manyhealth fields ranging from basic research tobedside care. The project is funded by a grantfrom the Division of Research Resources of theNational Institutes of Health (BiotechnologyResources Branch) for an initial term that ex-pires in July, 1978.



ARTIFICIAL INTELLIGENCE DENDRAL

The term "artificial intelligence" (AI) refersto research efforts aimed at studying and mecha-nizing information processing tasks that general-ly have been considered to require human intel-ligence. The current emphasis in the field is tounderstand the underlying principles in efficientacquisition and utilization of material know-ledge and representation of conceptual abstrac-tions in reasoning, deductive and problem-solving activities. AI systems are characterizedby complex computational processes that areprimarily non-numeric, e.g., graph-searching andsymbolic pattern analysis. They involve proce-dures whose execution is controlled by differenttypes and forms of knowledge about a given taskdomain, such as models, fragments of "advice"and systems of constraints or heuristic rules.Unlike conventional algorithms commonly basedon a well-tailored method for a given task, AIprocedures typically use a multiplicity of meth-ods in a highly conditional manner—depending onthe specific data in the task and a variety ofsources of relevant information. The tangibleobjective of this approach is the production ofcomputer programs which, using formal andinformal knowledge together with mechanizedhypothesis formation and problem-solving pro-cedures, will be more general and effective con-sultative tools for the clinician and medicalscientist.


This process, as implemented in the com-puter, is a simplified example of the cycle ofinductive hypothesis—deductive verification thatis often taught as a model of the scientificmethod (Whether this is a faithful descriptionof contemporary science is arguable, and how itmay be implemented in the human brain isunknown. Regardless, these are useful leadsrather than absolute preconditions for the prag-matic improvement of mechanized intelligencefor more efficient problem-solving.). The elabo-ration of these approaches with existing hard-ware and software technologies is the mostpromising approach to enhancing computerapplication to the vaguely structured problemsthat dominate our task domains.


*









Dr. Stanley Cohen, Associate Professor andHead of the Division of Clinical Pharmacology atStanford, directs this research in collaborationwith Dr. Stanton Axline and with computerscientists interested in artificial intelligence andmedical computing. An evolving computer pro-gram developed to assist physician nonspecialistsin the selection of therapy for patients withbacterial infections, MYCIN attempts to modelthe decision processes of medical experts. Itconsists of three closely integrated components:the Consultation System asks questions, makesconclusions and gives advice; the ExplanationSystem answers questions from the user tojustify the program's advice and explain itsmethods; and the Rule-Acquisition System per-mits the user to teach the system new decisionrules or to alter pre-existing rules judged to beinadequate or incorrect. Goals for further devel-opment of the system include expansion of theconsultation program to deal with infections

I other than bacteremias and implementation andJ evaluation of the system in the clinical setting atl Stanford University Hospital.



crystallographers at the University of California,San Diego. They are using the SUMEX-AIMfacility as the central repository for programs,data and other information of common interest.The general objective of the project is to applyproblem-solving techniques, which have emergedfrom artificial intelligence research, to the well-known "phase problem" of x-ray crystallo-graphy in order to determine the three-dimen-sional structures of proteins. The work is in-tended to be of both practical and theoreticalvalue to computer science (particularly artificialintelligence research) and protein crystallo-graphy.


A significant part of the SUMEX-AIM experi-ment is the development of a managementstruc-ture to maximize the utility of the computercapability for a national community.


DIALOG






t>

rithms.



proposed research.2) The relevance of the research to theartifici-





FACILITY INFORMATIONThe computer facility, consisting of a DEC

Model KI-10 CPU running under the TENEXoperating system, has 256K words (36-bit) ofhigh-speed memory, 1.6M words of swappingstorage, 70M words of disk storage, two 9-track800 bpi industry-compatible tape units, a dualDEC-tape unit, a line printer, and communica-tions-network interfaces providing user terminalaccess. SUMEX is available through TYMNETand as a host over the ARPANET communica-tions network.

Program (software) support will evolve fromthe basic system as dictated by the researchgoals and needs of the user. Initially, availableprograms include a variety of TENEX user,utility and text editor programs. Major userlanguages include INTERLISP, SNOBOL.SAIL,FORTRAN-10, BLISS-10, BASIC, Macro-10,OMNIGRAPHandMLAB.

RevisedMay 1975

POTENTIAL USERS




Biotechnology Resources BranchDivision of Research ResourcesNational Institutes of HealthBuilding 31, Room 58199000 Rockville PikeBethesda, Maryland 20014

-

TIhe Stanford University Medical Experi-mental Computer (SUMEX) was established inJanuary, 1974, to provide the first shared na-tional computing facility for medical research.Directed by Dr. Joshua Lederberg, Professor andChairman of the Department of Genetics,SUMEX is an innovative effort to help biomedi-cal scientists meet today's research requirementsand to explore computer applications in manyhealth fields ranging from basic research tobedside care. The project is funded by a grantfrom the Division of Research Resources of theNational Institutes of Health (BiotechnologyResources Branch) for an initial term that ex-pires in July, 1978.



ARTIFICIAL INTELLIGENCE

The term "artificial intelligence" (AI) refersto research efforts aimed at studying and mecha-nizing information processing tasks that general-ly have been considered to require human intel-ligence. The current emphasis in the field is tounderstand the underlying principles in efficientacquisition and utilization of material know-ledge and representation of conceptual abstrac-tions in reasoning, deductive and problem-solving activities. AI systems are characterizedby complex computational processes that areprimarily non-numeric, e.g., graph-searching andsymbolic pattern analysis. They involve proce-dures whose execution is controlled by differenttypes and forms of knowledge about a given taskdomain, such as models, fragments of "advice"and systems of constraints or heuristic rules.Unlike conventional algorithms commonly basedon a well-tailored method for a given task, AIprocedures typically use a multiplicity of meth-ods in a highly conditional manner— dependingonthe specific data in the task and a variety ofsources of relevant information. The tangibleobjective of this approach is the production ofcomputer programs which, using formal andinformal knowledge together with mechanizedhypothesis formation and problem-solving pro-cedures, will be more general and effective con-sultative tools for the clinician and medicalscientist.


DENDRAL


This process, as implemented in the com-puter, is a simplified example of the cycle ofinductive hypothesis—deductive verification thatis often taught as a model of the scientificmethod (Whether this is a faithful descriptionof contemporary science is arguable, and how itmay be implemented in the human brain isunknown. Regardless, these are useful leadsrather than absolute preconditions for the prag-matic improvement of mechanized intelligencefor more efficient problem-solving.). The elabo-ration of these approaches with existing hard-ware and software technologies is the mostpromising approach to enhancing computerapplication to the vaguely structured problemsthat dominate our task domains.









Dr. Stanley Cohen, Associate Professor andHead of the Division of Clinical Pharmacology atStanford, directs this research in collaborationwith Dr. Stanton Axline and with computerscientists interested in artificial intelligence andmedical computing. An evolving computer pro-gram developed to assist physician nonspecialistsin the selection of therapy for patients withbacterial infections, MYCIN attempts to modelthe decision processes of medical experts. Itconsists of three closely integrated components:the Consultation System asks questions, makesconclusions and gives advice; the ExplanationSystem answers questions from the user tojustify the program's advice and explain itsmethods; and the Rule-Acquisition System per-mits the user to teach the system new decisionrules or to alter pre-existing rules judged to beinadequate or incorrect. Goals for further devel-opment of the system include expansion of theconsultation program to deal with infections. other than bacteremias and implementation andevaluation of the system in the clinical setting atStanford University Hospital.



crystallographers at the University of California,San Diego. They are using the SUMEX-AIMfacility as the central repository for programs,data and other information of common interest.The general objective of the project is to applyproblem-solving techniques, which have emergedfrom artificial intelligence research, to the well-known "phase problem" of x-ray crystallo-graphy in order to determine the three-dimen-sional structures of proteins. The work is in-tended to be of both practical and theoreticalvalue to computer science (particularly artificialintelligence research) and protein crystallo-graphy.


A significant part of the SUMEX-AIM experi-ment is the development of a management struc-ture to maximize the utility of the computercapability for a national community.


DIALOG






.

rithms.



proposed research.2) The relevance of the research to the artifici-





FACILITY INFORMATIONThe computer facility, consisting of a DEC

Model KI-10 CPU running under the TENEXoperating system, has 256K words (36-bit) ofhigh-speed memory, 1 .6M words of swappingstorage, 70M words of disk storage, two 9-track800 bpi industry-compatible tape units, a dualDEC-tape unit, a line printer, and communica-tions-network interfaces providing user terminalaccess. SUMEX is available through TYMNETand as a host over the ARPANET communica-tions network.

Program (software) support will evolve fromthe basic system as dictated by the researchgoals and needs of the user. Initially, availableprograms include a variety of TENEX user,utility and text editor programs. Major userlanguages include INTERLISP, SNOBOL,SAIL,FORTRAN-10, BLISS-10, BASIC, Macro-10,OMNIGRAPHand MLAB.

Revised May 1975

;

<

POTENTIAL USERS




Biotechnology Resources BranchDivision of Research ResourcesNational Institutes of HealthBuilding 31, R00m58199000 Rockville PikeBethesda, Maryland 20014

Joshua - Stanford Universityfb827dx0405/fb827dx0405.pdf · Kraut, Dr. Stephan Freer and other...

Documents

Transcript of Joshua - Stanford Universityfb827dx0405/fb827dx0405.pdf · Kraut, Dr. Stephan Freer and other...