|
|
(Pavilion Suite 2)
Overview: 0. Why are new methods in epidemiology important? (Stacey/Majowicz) - public health, animal health, bioterrorism 1. Introduction to Epidemiology (Majowicz) 1.1) Background a) Definition of epidemiology and scope of the discipline, with examples b) Historical studies and development of analytic techniques / how epidemiologists have often incorporated techniques from other disciplines. c) How epidemiology relates to clinical medicine (where there has been use of ANNs more so than in epidemiology) 1.2) Modern Epidemiology a) Typical research questions and areas b) Typical study designs and data sets/formats 1.3) Current Analytic Techniques a) Descriptive analyses and data mining b) Statistical techniques (regression analyses, survival analyses, time series analyses) 2. Epidemiological Data (Stacey) 2.1 How is epidemiology data collected? 2.2 Major categories of data collection: a) survey data b) surveillance data c) public health data d) simulation data 2.3 Challenges to using epidemiology data: a) collection b) pre-processing c) missing values d) normalization e) data mining 3. Major Challenges in Epidemiology and Approaching the Natural Computation Approach (Majowicz) 3.1) Scientific challenges 3.2) Other challenges a) Data sharing and confidentiality b) Merging and using data from multiple, sometimes disparate sources (e.g. environmental monitoring data and administrative health data) 3.3) With respect to ANNs a) Defining how ANNs relate to "traditional" epidemiological analytic techniques Part II - A Natural Computation Approach 4. Supervised Methodologies (Stacey) 4.1 Classification using Artificial Neural Networks a) Use of Supervised ANNs to classify labeled and unlabeled data b) Interpreting model co-efficients; extracting meaningful population-level estimators from neural networks 4.2 Evolutionary Computation Techniques 5. Unsupervised Methodologies (Stacey) 5.1 Clustering with Unsupervised ANNs including SOMs, ART, etc. 5.2 Cluster Validation 5.3 Applications of clustering to case definition construction and evaluation 6. Future Directions (Stacey/Majowicz) 6.1 Epidemiology in general 6.2 Challenges to the introduction of natural computation techniques to epidemiology 6.3 Integration of standard statistical techniques with natural computation - a systems approach 6.4 Other considerations (e.g. accessing data, data mining and large data sets) 6.5 Infectious Disease Epidemiology and Enteric Disease Epidemiology in particular 6.6 Why all of this matters revisited - public health, animal health, bioterrorism a) Scientific considerations (e.g. bioterrorism, SARS and other emerging diseases)
Instructors' Background: Deborah Stacey is an Associate Professor of Computing and Information Science at the University of Guelph and a member of the Guelph Natural Computation Research Group. Her PhD is in Systems Design Engineering from the University of Waterloo where she worked in the Pattern Analysis and Machine Intelligence Lab. Currently her research group is examining techniques to improve a classification system's ability to use unlabeled data. This research builds on previous work with hybrid systems of supervised and unsupervised artificial neural networks and genetic algorithms. The computational demand of the research has resulted in work on developing parallel/distributed models of computation on cluster computers and using grid technologies. Dr. Stacey is a founding executive member of SHARCNET - a consortium of several universities in south-western Ontario that provides high performance computation for science, mathematics and business research. Shannon Majowicz completed her M.Sc. in Epidemiology (University of Guelph) in 1999, where her research project involved a descriptive analysis of human cryptosporidiosis cases reported in Ontario, Canada. Upon completion of her M.Sc., she joined the Foodborne, Waterborne and Zoonotic Infections Division, Health Canada, as an epidemiologist with the Division's waterborne diseases group, working on a strategy for linking environmental and water-related risk factor data to human enteric and waterborne disease surveillance data using Geographical Information System applications. Her current activities include the co-ordination of the Division's National Studies on Acute Gastrointestinal Illness (NSAGI) initiative, which consists of a series of inter-related studies that together will provide the first population-level data on gastrointestinal illness in Canada. Shannon is also a PhD candidate in the Department of Population Medicine at the University of Guelph, where her thesis is an evaluation of the magnitude, distribution, and impact of self-reported, acute gastrointestinal illness in the community. Her research interests include evaluating the impact of case definitions on infectious disease surveillance and research, and exploring the application of various analytic techniques including artificial neural networks and fuzzy logic to epidemiologic studies.
|