Dr. Hamish S F Fraser
Curriculum Vitae June 1999
PERSONAL DETAILS
Born: May 19th 1962
Nationality: British
Address: 16 Hill Street, Somerville, MA, 02144, USA.
Telephone, work: 617 258 8996
Email: hamish@medg.lcs.mit.edu
Home Page: http://medg.lcs.mit.edu/hamish/home.htm
MAIN QUALIFICATIONS
1999: Registered for MD thesis at Edinburgh University (write-up in progress)
1998: USMLE parts I, II and III (1996 - 1998), Massachusetts Medical Licence, 1998
1991: M.Sc Knowledge Based Systems, Edinburgh University
1990: MRCP Edinburgh (medical boards equivalent)
1986: MB ChB, Edinburgh University (medical degree)
1984: B.Sc. Med. Sci/Physiology, Edinburgh University
EDUCATION
1973 - 74: Dr Challoner's Grammar School, Amersham, Bucks, England.
1974 - 79: The Marr College, Troon, Ayrshire, Scotland.
1979 - 80: Langside College, Prospecthill Road, Glasgow, Scotland.
School qualifications
Highers: English, Maths, Biology, Physics, Chemistry, Geography (all A)
A Levels: Physics: A, Chemistry: A, Biology: B.
Prizes: Science: 1975 ,Chemistry: 1976, Physics: 1977
Edinburgh University Medical School 1980-1986
Dissertation: ``The Comparative Physiology of the Basal Ganglia.''
CURRENT POST
September 1995 to the present,
Research fellow in the Divisions of Clinical Decision Making and Cardiology, Tufts-New England Medical Center (50%)
Research affiliate in the Clinical Decision Making Group at the Laboratory for Computer Science of the Massachusetts Institute of Technology (50%).
I have spent three years working on the Heart Disease Program, a large expert system (based on a modified Bayesian Belief Network) developed by Bill Long to assist with the diagnosis of heart disease. During that time we developed a new user interface and have increased the knowledge base considerably, adding new diseases from cardiology and general medicine in the light of the evaluation work.
MD Project Work
I have designed, set up and performed a clinical trial of the Heart Disease Program with physicians (mainly residents) at the New England Medical Center. We developed the user interface as a Web site to allow physicians to access the program over the Internet. 127 cases have been directly entered by participating physicians and 50 entered over the Internet from around the world, many by cardiologists. This work has been presented at the American Medical Informatics Association, the American College of Cardiology and the British Cardiac Society meetings. The work is forming the basis for an MD thesis submission to Edinburgh University.
Machine Learning/Data Analysis
My second major area of work has been in using tools from machine learning and statistics to develop decision support tools from data. I started this area of work with Dr. Lee Kennedy in Edinburgh, developing models to predict the diagnosis of myocardial infarction in patients with chest pain. I continued this work with Bill Long and Chris Tsien at MIT exploring ways to optimise decision trees induction, improve generalisation of models over data collected in different countries and to deal with missing data.
In 1997/98 I received funding from Pfizer Inc. to develop models that can predict adverse events in clinical trials. Techniques applied to this problem included machine learning, logistic regression and backpropagation neural networks. I continue to be interested in the detection of trends over time, and the use of patterns in medication to predict adverse events.
Patient Directed Decision Support Tools
My most recent grant is from the National Heart Attack Alert Program, a group funded by the NIH to explore ways to assist patients with myocardial infarction to recognise and act on their symptoms early. I am assessing the performance of current patient guidelines on data from Edinburgh and Boston. I am running a randomised trial to assess the effect of patient educational materials on the ability of patients to recognise and act on symptoms of MI. In addition we are exploring the use of the Internet and palm top computers to provide teaching and advice to patients with heart disease.
MEDICAL POSTS HELD
Registrar Training In Cardiology: NTN(R) held in Edinburgh
October 1992 to July 1995: Career Registrar in Cardiology
Edinburgh Royal Infirmary, 1 Lauriston Place, Edinburgh, Scotland.
and Falkirk and District Royal Infirmary, Majors Loan, Falkirk, Scotland.
Experience:
Cardiology, Edinburgh: Dr. Nicolas Boon and Dr. Hugh Miller
22 Months as a specialist cardiology registrar.
General Medicine, Falkirk: Dr. Stanley Wright and Dr. William Ruddell
1 year of very busy general medicine with emphasis on respiratory medicine and gastroenterology and initial training in echocardiography.
February to July 1992: Locum Registrar in Cardiology (5+ months)
The Freeman Hospital, Freeman Road, High Heaton, Newcastle.
Dr. Roderick Bexton and Dr. Janet McCoomb. general and invasive cardiology.
Senior House Officer Posts
August 1989 to August 1990: Senior House Officer in Cardiology and Medicine
The Northern General Hospital, Sheffield, UK.
General Medicine: (6 months) under Professor D. Munro, with specialist interests in endocrinology and connective tissue disease.
Cardiology: (6 months) under Dr. David Oakley, general and invasive cardiology.
August 1988 to July 1989: Senior House Officer in Medicine
Stirling Royal Infirmary, Livilands Gate, Stirling, Scotland.
Under Dr. Sheila Reith and Dr. A. Clark. Experience in all areas of general medicine with particular emphasis on diabetes and endocrinology and gastroenterology.
August 1987 to July 1988: Senior House Officer in Medicine/Geriatrics
Falkirk and District Royal Infirmary, Majors Loan, Falkirk, Scotland.
Dr. William Ruddell, and Dr. John Miller
A busy post in general and geriatric medicine.
House Officer Posts
August 1986 to February 1987, Surgery.
Department of Clinical Neuro-science, Western General Hospital, and
Ward 20, The Royal Infirmary, Edinburgh
February 1987 to August 1987, General Medicine Wards 28 and 31,
The Royal Infirmary, Edinburgh.
MEDICAL EXPERIENCE
General Medicine
3 years general professional training as SHO. 1 year general medicine as a registrar. Specialist units included diabetes and endocrinology, respiratory medicine, gastroenterology, neurology and geriatric medicine.
Procedures
Cardiac catheterisation: 300 + Left, 50 Right (Sheffield, Newcastle, Edinburgh)
Pacemaker implantation: 60 +
Invasive electro-physiology: 10+ procedures
Echocardiography: reporting 20+ per week for 18 months, 200+ personally performed
12 lead ECG, 24hr tape and ECG stress rest reporting, extensive experience.
Other procedures: chest drains, lumbar puncture, central lines, pleural drainage, sigmoidoscopy. ACLS certified.
Current Clinical Involvement
RESEARCH TRAINING
1990 - 91: M.Sc. in Knowledge Based Systems (Artificial Intelligence Department)
This was a one-year M.Sc., comprising a 6-month project and the following courses:
Project and Thesis: "Qualitative Modelling of the Heart and Circulation"
Supervisors: Dr. P. Ross, R. Weiss, Dr N. Boon
This project involved building a large computer program that enabled qualitative models of physical or biological systems to be created and run. Several models of cardiac physiology were built. The modelling system was programmed in Prolog and based on QSIM (developed by Ben Kuipers).
Courses at MIT (full semester, audited):
1995: Structure and Interpretation of Computer Programs
1996: Knowledge Based Systems
1996: Introduction to Interactive Programming - user interfaces and networks programming taught in Java
1997: Medical Computing
1998: Machine Learning
Other courses:
1996: Regression Analysis - 2 week intensive course taught by Stanley Lemeshow (Tufts)
1996: Security and Confidentiality of Medical Records, a two-day conference run by the Harvard Community Health Plan. (Harvard)
1996: The Electronic Medical Record, a two-day course run by the Center for Clinical Computing at the Beth Israel Hospital, Boston (Harvard)
1997: IBM sponsored weekend course in Java programming (MIT)
1997: Echocardiography: Reading with the Experts, two day course run by Dr. Pandian from the New England Medical Center (Tufts)
1997: The Design Team of the Future, course on the impact of new technologies on design (MIT Sloan Management School)
1998: Design of epidemiological studies and epidemiological data analysis, two week course taught by Kenneth Rothman (Tufts)
CONFERENCE PRESENTATIONS AND POSTERS:
Evaluation of Guidelines to Help Patients Recognise Myocardial Infarction
Hamish S F Fraser, Lucila Ohno-Machado, Vincent Carey, Greg Sharp, Shapur Naimi,
To be presented as a poster at the AMIA Fall Symposium on Computer Applications in Health Care, Washington DC, November 1999.
Evaluating A Decision Support System Using the Web
Hamish S. F. Fraser, William J. Long, Shapur Naimi
To be presented as a poster at the AMIA Fall Symposium on Computer Applications in Health Care, Washington DC, November 1999.
Lower use of Thrombolysis in Women is not due to Bias in Initial Assessment
H S F Fraser, L M Smeaton, S Naimi
Presented at the British Cardiac Society Annual Meeting, Manchester, May 27th 1999
Interfacing the Heart Disease Program to Coded Data
William J Long, Hamish S F Fraser, Shapur Naimi.
Presented at the AMIA Fall Symposium on Computer Applications in Health Care, Orlando, Florida, November 1998.
Automating The Coding Of Medical Events In Multiple Dictionaries
Long, WJ, Fraser HSF, Trost, C
Poster presentation, Society for Medical Decision Making Annual Meeting, Cambridge, MA, October 1998
Prospective Clinical Evaluation of a Computer Program for Cardiac Diagnosis with Particular Focus on More Serious Diseases
Hamish S F Fraser, William J Long, Shapur Naimi.
Presented at the British Cardiac Society Annual Meeting, Glasgow, Scotland, May 1998.
Prospective Clinical Evaluation of a Computer Program to Assist with the Diagnosis of Cardiac Disease
Hamish S F Fraser, William J Long, Shapur Naimi
Presented as a poster and demonstration at the American College of Cardiology Annual Meeting, March 1998, Atlanta GA.
Optimising Diagnosis of Myocardial Infarction in the Emergency Room: A new Flowchart Decision Aid
Christine L Tsien, Hamish S F Fraser
Presented at the American College of Cardiology Dearborn Summit:
Reducing Costs and Improving Performance in Cardiovascular Care-Practical Lessons
Sept. 26 - 27, 1997, program director Kim A. Eagle, MD, FACC.
Comparing Complex Diagnoses: A Formative Evaluation of the Heart Disease Program
Hamish S F Fraser, William J Long, Shapur Naimi
Poster presentation at the 1997 AMIA Fall Symposium on Computer Applications in Health Care. Nashville, Tennessee, USA. 26 – 29th October 1997.
Collecting and Interpreting Temporal Data in an Expert System
William J. Long, Hamish Fraser, Shapur Naimi, James Stahl
Poster presentation at the 1997 AMIA Fall Symposium on Computer Applications in Health Care. Nashville, Tennessee, USA. 26 – 29th October 1997.
Testing a Heart Disease Diagnosis Program in a Practical Clinical Setting.
Experience in Developing a Web Interface to allow Physicians in a General Medical Clinic to use a Large Diagnosis Program.
Hamish S F Fraser, William J Long, Shapur Naimi
Presented to the AAAI symposium on Artificial Intelligence Applications in Health Care, Stanford, March 1996.
Can Cardiac Marker Protein Measurements at Presentation Improve Early Diagnoses of Myocardial Infarction?
R. Lee Kennedy, A M Burton, W G Hamer, L N McStay, J McIntyre, H S F Fraser, K A A Fox, R F Harrison, European Cardiac Society September 1995.
A Simple Statistical Model to Predict Complications and Death in Patients Admitted to Coronary Care Unit.
Kennedy R.L., Harrison R.F., Fraser H.S., Burton A, McStay L, Fox K.A.A., Woods K.L. European Cardiac Society, September 1995
Development of an Artificial Neural Network Trained to Diagnose Myocardial Infarction from Clinical and ECG Data.
Kennedy R.L., Fraser H.S., Burton A, McStay L, Harrison R.F., Steedman D.J., Fox K.A.A. European Cardiac Society, September 1995
Logistic Regression (LR) and Neural Network (NN) Models for Early Diagnosis of AMI.
Kennedy R.L., Burton A, Fraser H.S., McStay L, Harrison
R.F., Fox K.A.A. British Cardiac Society, May 1995.
Review of Triage, Diagnosis and Readmission Rates in Patients with Acute Chest Pain
Kennedy R.L., Burton A, Fraser H.S., McStay L, Steedman D.J.,
Fox K.A.A. British Cardiac Society, May 1995.
A Comparison of Radial Basis Functions and Back-Propagation in the Diagnosis of Myocardial Infarction.
Hamish S F Fraser, R. Lee Kennedy, Peter Ross, Robert Harrison.
Presented at the International Conference of Expert Systems and Neural Networks in Medicine, Plymouth, UK, August 1994.
SHORT PAPERS WITH PRESENTATIONS
Differential Diagnoses of the Heart Disease Program have better Sensitivity than Resident Physicians.
Hamish S F Fraser, William J Long, Shapur Naimi.
Presented at the AMIA Fall Symposium on Computer Applications in Health Care, Orlando, Florida, November 1998.
Modelling Treatment of Ischemic Heart Disease with Partially Observable Markov Decision Processes.
Milos Hauskrecht, Hamish Fraser
Presented at the AMIA Fall Symposium on Computer Applications in Health Care, Orlando, Florida, November 1998.
Improving Machine Learning Performance by Removing Redundant Cases in Medical Data Sets
L. Ohno-Machado, H.S. Fraser, A. Ohrn
Presented at the AMIA Fall Symposium on Computer Applications in Health Care, Orlando, Florida, November 1998.
Using Classification Tree and Logistic Regression Methods to Diagnose Myocardial Infarction
Christine L Tsien, Hamish S F Fraser, William J Long, R Lee Kennedy
Presented at MEDINFO 98, Seoul, Korea, August 1998.
LRTree: a hybrid technique for classifying myocardial infarction data containing unknown attribute values.
Christine L Tsien, Hamish S F Fraser, Isaac S Kohane
Tsien CL, Fraser HS, Kohane IS. In: Wu X, Kotagiri R, Korb KB, eds., Research and
Development in Knowledge Discovery and Data Mining. Lecture Notes in Artificial Intelligence 1394. Berlin Heidelberg: Springer-Verlag 1998: 409-411.
Planning Medical Therapy Using Partially Observable Markov Decision Processes
Milos Hauskrecht, Hamish Fraser
Presented at Dx98, Cape Cod, May 1998
A Web Interface for the Heart Disease Program
William J. Long, Hamish Fraser, Shapur Naimi
Presented at the American Medical Informatics Association Fall Symposium, 1996
JOURNAL ARTICLES
Using the Technology of the World Wide Web to Manage Clinical Information
Hamish S F Fraser, Isaac S Kohane, William J Long
British Medical Journal, No 7094 Volume 314 Saturday 31 May 1997 P 1600-4
Full text of the article on the BMJ website:
http://www.bmj.com/archive/7094ip1.htm
Reasoning Requirements for Diagnosis of Heart Disease
W. J. Long, H. Fraser, and S. Naimi,
Artificial Intelligence in Medicine 10 (1997) pp. 5 - 24
An Artificial Neural Network System For Diagnosis Of Acute Myocardial Infarction (AMI) In The Accident And Emergency Department: Evaluation And Comparison With Serum Myoglobin Measurements.
Kennedy RL, Harrison RF, Burton AM, Fraser HS, Hamer WG, MacArthur D, McAllum R,
Steedman DJ,
Comput Methods Programs Biomed 1997 Feb;52(2):93-103
Early Diagnosis of Acute Myocardial Infarction Using Clinical and Electrocardiographic Data at Presentation: Derivation and Evaluation of Logistic Regression Models
R.L. Kennedy, A.M. Burton, H.S. Fraser, L.N. McStay, R.F. Harrison
European Heart Journal, Vol. 17, August 1996, p1181 - 91
Medinfo-92 Conference Report
H S F Fraser
The Lancet, vol. 340: Sept. 26, 1992, pp 784.
SUBMITTED JOURNAL ARTICLES
Using Patient-Reportable Clinical History Factors to Predict Myocardial Infarction
Samuel J. Wang, Lucila Ohno-Machado, Hamish S F Fraser.
Submitted to the Journal of the American Medical Informatics Association
Modeling Treatment of Ischemic Heart Disease with Partially Observable Markov Decision Processes.
Milos Hauskrecht, Hamish Fraser
Submitted to Artificial Intelligence in Medicine.
Evaluating a Cardiac Diagnosis Program in a Clinical Setting
Hamish S F Fraser, William J Long, Shapur Naimi.
Under submission to the British Medical Journal
INVITED TALKS AND TEACHING
Programming Skills
PROGRAMMING PROJECTS
Disease Program written in Java.
Consulting Work
Websites
http://www.nemc.org/smdm/ (now at http://www.gwu.edu/~smdm/)
http://medg.lcs.mit.edu/hamish/demo/vitals.htm
CAREER INTENTIONS
I intend pursuing a career in Medical Informatics/Artificial Intelligence with a particular emphasis on the application of these techniques to cardiology. Experience in the US suggests that researchers with both medical and computer science training are particularly productive in this field. Having completed the majority of my clinical training in cardiology and with experience and qualifications in AI/computing techniques, I am keen to continue working in both Informatics and Cardiology. I will submit this work for a British MD thesis (similar to a PhD) at Edinburgh University, which can be based on work performed abroad.
OUTSIDE INTERESTS
Photography, travel (35 countries so far...), Scuba diving, sailing, hill walking, snow shoeing conservation, cooking Scottish food including (the first?) kosher haggis, swing dancing, singing and a variety of ethnic music and dances, not all of them Scottish.