perm filename NLM[AM,DBL]1 blob
sn#196171 filedate 1976-01-13 generic text, type T, neo UTF8
00100 .sec OVERVIEW
00200
00300 In response to a stated need for bringing computer technology into
00400 medical education, and eventually into the delivery of health care,
00500 we propose to train current and future educators of medical students
00600 in the art and science of applying computers to medical problems.
00700 Stanford University already has many faculty and students actively
00800 working on cross-disciplinary problems of biomedical computing. We
00900 intend to exploit this active research and teaching environment in
01000 the service of teaching individuals about the techniques and problems
01100 of using computers in medicine.
01200
01300 .sec OBJECTIVES
01400
01500 We propose to train people who can integrate knowledge of computer
01600 science with medical and biomedical expertise by providing
01700 interdisciplinary classroom and research training in biomedical
01800 computing. Three educational opportunities will be made available to
01900 trainees: (i) special new seminars in biomedical computing, (ii)
02000 existing University courses, especially in the Computer Science Dept,
02100 and (iii) research projects that will provide supervision on
02200 individual pieces of research. (The environment in which these
02300 research projects operate is discussed in more detail below.)
02400
02500 <<Seminars>. Prof. Feigenbaum and Dr. Buchanan will lead a series of
02600 seminars on biomedical computing for the trainees. We will expect
02700 every trainee to participate in these seminars, mostly to teach them
02800 about the topics to be discussed, but also to give them a forum for
02900 common problems and a sense of community. Because of the wide
03000 variety of applications of computers in and around Stanford Medical
03100 School and the larger Stanford community, a number of additional
03200 faculty will discuss their own research and their own views of
03300 computers in biomedical work.
03400
03500 <<Courses>. The courses routinely offered by the COMPUTER SCIENCE
03600 DEPARTMENT will be available to the trainees. Trainees will be
03700 expected to take about two courses each academic quarter, the number
03800 and subjects depending on their own computing background. Trainees
03900 will take courses in programming, artificial intelligence, data
04000 structures, computer design and computer systems. Non-clinical
04100 medical school courses, such as biostatistics or computing seminars
04200 offered by various departments (last year in ***) will be available.
04300 Additional courses outside the COMPUTER SCIENCE DEPARTMENT and the
04400 Medical School that would interest some trainees would also be open
04500 to them, such as courses in statistics and operations research.
04600
04700 <<Research>. An important requirement on the trainees will be
04800 individual research under the supervision of one of the affiliated
04900 faculty listed below. We strongly believe that first-hand, specific
05000 knowledge of computing problems in one medical area is a necessary
05100 complement to the general knowledge gained from course work. The
05200 nature of an individual's research project will naturally depend on
05300 his/her interests, experience and degree goals, but will involve
05400 writing programs and working closely with people who design and write
05500 them. The biomedical computing research environment at Stanford is
05600 unique. Because of the NIH-sponsored SUMEX/AIM computing resource,
05700 several research projects around the country are "housed" at this
05800 local facility. In addition, the training faculty are actively
05900 involved in several local projects.
06000
06100 In summary, our objectives are to train people in biomedical
06200 computing by providing:
06300
06400 1) training in computer science to medical faculty and potential
06500 teachers of medical students, including potential computer science
06600 faculty members whose career goals will take them into medical school
06700 environments;
06800
06900 2) a research environment and supervision for individual research
07000 projects that will integrate computer science techniques, especially
07100 AI techniques, with medical and biomedical problems.
07200
07300 .sec RESOURCES
07400
07500 .ss Computer Science Department
07600
07702 The Computer Science Department aims to acquaint students with the
07704 technological and intellectual roles of automatic digital computers,
07706 and to educate research workers in computer science. In spite of
07708 the diversity of the applications, the methods of attacking problems
07710 with computers show a considerable unity, and computer science is
07712 concerned with the underlying principles. The field is still young,
07714 and the student will find many more questions than answers.
07716
07718 The Department has competence in artificial intelligence, combinatorics
07720 and analysis of algorithms, design of computer and program systems,
07722 mathematical theory of computation, and numerical analysis.
07724
07726 Since computer science is inherently interdisciplinary, graduate
07728 students of computer science are encouraged to include in their study
07730 program a good deal of work in other departments. The CSD combines
07732 a well-respected faculty, a diverse program of courses and
07734 sophisticated computing facilities to offer a high-quality research
07736 and educational environment to its students.
07800
07900 .ss On-Going Research Projects
08000
08100 Several research projects at Stanford provide a rich environment for
08200 individual research as well as illustrative problems for discussion.
08300 Projects with which the Principal investigators are actively involved
08400 as part of (or overlapping with) the Heuristic Programming Project
08500 are (i) MYCIN, (ii) Heuristic DENDRAL, (iii) X-Ray Crystallography,
08600 (iv) MOLGEN. These are discussed in separate appendices.
08700
08800 Numerous other projects around the Medical School make sophisticated
08900 applications of computers for data acquisition and analysis, e.g.,
09000 the drug interaction warning system in the hospital pharmacy, a
09100 patient monitoring system in the intensive care unit of the
09200 Cardiology ward of the hospital, and a premature infant screening
09300 system in the Genetics Research Center and Pediatrics.
09400
09500 Joint efforts by the Heuristic Programming and MYCIN projects have
09600 resulted in a "production-system" framework to facilitate developing
09700 more medical reasoning programs like MYCIN. Besides MYCIN's
09800 knowledge of bacteremia diagnosis and therapy we have encoded
09900 knowledge of meningitis diagnosis and therapy. Other research groups
10000 have used this framework to encode knowledge of venereal diseases
10100 (Univ. of West Virginia) and psychiatric disorders (Univ. of
10200 California at Irvine). These running programs are valuable teaching
10300 aids for trainees. This general framework will help trainees to
10400 develop medical reasoning programs in their own areas of special
10500 interest.
10600
10700 There are limitations on the framework, of course, and understanding
10800 these would be an important part of learning about it. One of its
10900 chief virtues is giving system designers the ability to separate
11000 medical knowledge from the computer program that uses it to reason
11100 about different problems.
11200
11300
11400 *** <Put descriptions of projects listed in separate Appendices> ***
11500
11600 .ss SUMEX/AIM Computer Facility
11700
11800 The National research resource for applications of AI to medicine
11900 (AIM) is based at Stanford, with Prof. Lederberg as PI.While the
12000 facility is available on a national basis to qualified investigators,
12100 local access to the professional and technical staff is an important
12200 asset for our trainees. The present project will qualify for access
12300 to SUMEX-AIM as a STanford subproject with respect to AI research by
12400 trainees.
12500
12600
12700 *** <more SUMEX description here> ***
12800
12900 .sec TRAINEES
13000
13100 There are four types of trainees with whom we have already had some
13200 training experience. We would continue soliciting applications from
13300 all four groups and tailor the training program to the backgrounds of
13400 the trainees. The four groups are discussed briefly in the next
13500 four subsections.
13600
13700 .ss Post-MD Clinical Fellows
13800
13900 The MYCIN project, under NIH funding, supports clinical fellows in
14000 Infectious Disease and Clinical Pharmacology. This project has not
14100 been funded to train physicians in computing, but has nevertheless
14200 given them a good enough background to initiate their own biomedical
14300 computing research or to teach about uses of computers in medicine.
14400
14500 Under the training program, we would extend the effort to strengthen
14600 the computing background of these clinical fellows by involving them
14700 primarily in the seminars . Their day to day contact with an
14800 advanced research project exposes them to the concepts covered in
14900 many of the CSD courses, so we would not insist that MD's spend much
15000 of their time in classes. We would expect the MD's not to be
15100 particularly interested in obtaining an advanced degree in Computer
15200 Science, and would make their training less dependent on course work.
15300
15400 .ss Medical Students Preparing for Academic Careers
15500
15600 Several medical students have inquired about the possibility of
15700 studying at Stanford for short periods of time while "on leave" from
15800 their medical schools. We have accommodated one student from the
15900 University of California medical school at Irvine for Fall Quarter,
16000 1975-76. All of these inquiries have been initiated by the students
16100 -- we have not solicited them, nor have we been able to offer
16200 financial support.
16300
16400 In one academic quarter it is only possible to provide a basic
16500 framework for thinking about biomedical computing and supervise a
16600 small programming (or other) research project. Under the training
16700 program, we would be able to provide financial support for students
16800 to allow them to spend as much time as they need to become effective
16900 teachers. In one year these students would be able to obtain a
17000 Master's degree in computer science if they wished, as described in
17100 the administrative section below. We would urge them to do so.
17200
17300 .ss Interdisciplinary PhD Students
17400
17500 Training interdisciplinary students is particularly appropriate for
17600 the goals of the NLM, especially when the research spans computer
17700 science and medicine, or computer science and one of the biological
17800 sciences. Stanford encourages such interdisciplinary work and thus
17900 attracts unique individuals even without solicitation. We would
18000 expect that the availability of training money for supporting these
18100 individuals would increase the attractiveness of Stanford.
18200
18300 The MYCIN project began as the PhD research project of one individual
18400 in an interdisciplinary program spanning Computer Science,
18500 Biostatistics and Clinical Pharmacology. In slightly more than two
18600 years, this person took enough courses and performed the research to
18700 satisfy the PhD requirements of the University.
18800
18900 .ss PhD Students in Computer Science Interested in Biomedical
19000 Computing
19100
19200 It is important to make potential Computer Science faculty aware of
19300 medical computing problems in order to establish bases in other
19400 universities for communication with their own medical and pre-medical
19500 students. We have found a great interest among students in the CSD
19600 in medically related problems, and we support several of them to do
19700 research on various aspects of these problems.
19800
19900 Under current Heuristic Programming and MYCIN Project research
20000 support, we are training seven PhD candidates in the Computer Science
20100 Dept. in medical computing. Five of these persons are working
20200 closely with the MYCIN project at present, although only one is
20300 supported by that project. Their research has dual aspects: on the
20400 one hand, it focuses on the fundamental problem of managing a large
20500 amount of medical knowledge in a complex program, and on the other
20600 hand, it aims for a high performance medical reasoning program. That
20700 is, the students already working with us are building a strong base
20800 for teaching careers that will intersect with medical teaching and
20900 research.
21000
21100 .sec TRAINING EXPERIENCE
21200
21300 As mentioned above, the Heuristic Programming and MYCIN Projects are
21400 already collaborating with and training several persons in much the
21500 manner envisioned by the NLM. In this section, we present brief
21600 sketches of some of these persons to indicate in more detail the
21700 extent of our involvement.
21800
21900 Physicians: (S.Axline, F.Rhame, V.Yu, J.Heiser)
22000
22100 Medical Students: (D. Ludwig, C.Risk)
22200
22300 Interdisciplinary PhD: (T.Shortliffe)
22400
22500 Computer Science Department PhD Candidates: (R. Davis, J. Aikins,
22600 B.Clancey, B. VanMelle, L. Fagan)
22700
22800
22900 .sec ADMINISTRATIVE DETAILS
23000
23100 .ss SUPERVISION
23200
23300 Trainees will be supervised by Prof. Feigenbaum, Dr. Buchanan, and R.
23400 Davis. This extends from normal advising to selection of suitable
23500 research projects, and includes helping with course selection and
23600 problems of finding one's way around Stanford.
23700
23800 .ss RECRUITMENT AND ADMISSIONS
23900
24000 To a small extent Stanford already attracts both medical and CS
24100 students who actively seek the opportunity to learn about, and do
24200 research in, medical computing. The Heuristic Programming, MYCIN,
24300 and SUMEX projects are known to many as the center of such work at
24400 Stanford. Thus we receive inquiries from more students and
24500 post-doctoral fellowship applicants than we can financially support.
24600 Upon starting the training program proposed here we would hope to
24700 respond favorable to the better candidates.
24800
24900 We will announce the availability of training positions throughout
25000 medical schools and computer science departments around the country.
25100 At professional meetings and speaking engagements we will arrange to
25200 talk with groups of interested people, much as we do now. We will
25300 interview candidates in person whenever possible, and by telephone
25400 otherwise.
25500
25600 For prospective trainees who seek a PhD degree, either in the CSD or
25700 in an interdisciplinary program, the CSD admissions committee will do
25800 the preliminary screening of candidates together with all other
25900 applicants to the department. The chairman of that admissions
26000 committee and Prof. Feigenbaum will make the final selection of the
26100 candidates to be admitted to this program.
26200
26300 .ss DEGREES
26400
26500 Trainees who seek a degree will be in one of three degree programs of
26600 the University.
26700
26800 The University's basic requirements for the doctorate (residence,
26802 dissertation and examination) will be satisfied for PhD candidates
26804 in the CSD. In addition the CSD requires that doctoral students
26806 pass a comprehensive exam covering introductory level graduate
26808 material in major areas of computer science and a qualifying
27100
27200 Candidates for an interdisciplinary PhD degree are first admitted to
27300 the CSD and then map out a special program of course work and
27400 research, supervised by an interdisciplinary faculty committee,
27500 culminating in a "special" area PhD.
27600
27700 The third degree program we will offer is more appropriate for
27800 persons with an MD degree who wish to establish some credentials in
27900 computer science without fulfilling all the requirements necessary
28000 for a PhD degree. This is a program leading to a Masters degree in
28100 Computer Science. The requirements for an MA degree will be tailored
28200 by the CSD to the trainees in this program. In paerticular, some of
28300 the courses normally required for MA candidates will be waived for
28400 trainees (computer hardware, numerical analysis, computability
28500 theory), leaving them free to substitute other courses. They will be
28600 expected to complete 42 course units, as are all MA candidates, but
28700 will have more flexibility. In addition, these persons will be given
28800 a specail examination for the MA degree covering the core CS areas
28900 (programming techniques, languages, sytems, and artificial
29000 intelligence), instead of all areas.
29100
29200
29300 .sec BUDGET
29400
29500 We request support for five years to cover expenses of the
29600 training program.
29700
29800 Personnel: EAF 20%
29900 BGB 20%
30000 25% of one FTE faculty member in Computer Science
30100 (to be split among R.Floyd, D.Knuth, G.Danzig,
30200 C.Green, T.Winograd -- with their consent)
30300
30400 Research Associate 100%
30500 Secretary 50%
30600 Visiting Lecturers (honoraria)
30700
30800 Trainees: 3 post-docs
30900 3 predoctoral students
31000
31100 Travel:
31200
31300 Terminals: Purchase 3 terminals (approx $10,000)
31400
31500 Computer Time: SUMEX-AIM (no charge for AI research work)
31600 SCIP (approx $6000 per year -- $500/mo.)
31700
31800 Office Expenses: Phone
31900 Supplies
32000
32100 Hardware Purchase: 256K memory for SUMEX (approx $100,000)
32200
32300 TOTAL: Approx. $210,000 in first year
32400 90,000 in years 2-5
32500
32600
32700
32800 .ss BUDGET NOTES
32900
33000 Computer Time. The STanford Center for Information Processing (SCIP)
33100 is the facility at which much biomedical computing is done at
33200 Stanford. In order to demonstrate these programs and interact with
33300 these research groups, the trainees will need to have access to the
33400 IBM 370 computer at SCIP. Money must be budgeted for terminal access,
33500 computer time and file storage on this machine.
33600
33700
33800 Travel. Travel money is to be used for three purposes: to send
33900 trainees to conferences, to bring visiting lecturers to Stanford for
34000 this program, and to provide for interviewing and recruiting
34100 prospective trainees.
34200
34300 Terminals. We will need to provide terminals for access to the
34400 SUMEX-AIM and SCIP computers. Purchase price is almost exactly the
34500 same as two years' leasing cost, so we have requested money to
34600 purchase the three terminals we will need. (2 Datamedias, 1 TI -- ?)
34700
34800 Personnel. The CSD faculty listed will contribute a small, but
34900 significant, fraction of their time to training the students and
35000 post-doctoral fellows supported by this program. They are listed
35100 together under one quarter-time equivalent in the CSD to simplify
35200 budgeting and accounting.
35300
35400 Office Expenses. The most significant expense is for computer
35500 terminal paper (for the TI terminal). We estimate using *** rolls of
35600 heat-sensitive paper per month at *** per roll. Phone and supplies
35700 are estimated to be 20% of the Heuristic Programming Project totals.
35800