The IEEE Ottawa Section, IEEE Ottawa Consultants Network (AICN), and Engineering in Medicine and Biology Society (EMBS) invites all interested IEEE, IET members and other engineers, technologists, and students to a technical presentation on:
Successfully held: Wednesday April 30, 2014
Knowledge Discovery and Data Mining (KDD) is the nontrivial process of extracting implicit, novel, and useful information from large volume of data. A multi-disciplinary field of science and technology, KDD includes statistics, database systems, computer programming, machine learning, and artificial intelligence. It spans a wide range of applications in Engineering (intrusion detection and network security, flow classification), business (fraud detection, decision support systems, forecasting market trend), medicine and population health (study of drug implications, disease outbreak), and environmental science (flood prediction).
Major applications of Knowledge Discovery and Data Mining in healthcare fall into four categories: (a) Clinical Medicine: Modern hospitals and clinical centers surpassed their traditional role as a place for diseases’ diagnosis and treatment and now acting as a mass database and a source of complex clinical, laboratory, equipment use, and drug management data which can be analyzed for disease diagnosis and decision making; (b) Public Health: including early outbreak detection, healthcare and syndromic surveillance; (c) Healthcare Text mining: including mining medical literature, as well as mining clinical data such as patients’ clinical records; and (d) Healthcare Policy and Planning: including detecting expensive clinical profiles among patients diagnosed with a specific chronic illness which has a high disease’s burden such as diabetes. Data mining also helps health planners to solve resource allocation problems and capacity issues.
In this talk, we present the results of two recent studies conducted in the Knowledge Discovery and Data Mining lab at the University of Ottawa:
(a) Predicting High Cost Patients in General Population using Data Mining Techniques
We applied data mining techniques to a nationally-representative expenditure data from the US to predict very high-cost patients in the top 5 cost percentiles, among the general population. A dataset of 100,000 records was pre-processed and modeled by Decision Trees (including C5.0 and CHAID), and Neural Networks. Multiple predictive models were built and their performances were analyzed using various measures including correctness accuracy, and G-mean. We concluded that among a primary set of 66 attributes, the best predictors to estimate the top 5% high-cost population include individual’s overall health perception, history of blood cholesterol check, history of physical/sensory/mental limitations, age, and history of colonic prevention measures. We predicted high-cost patients without knowing how many times the patient was visited by doctors or hospitalized. Consequently, the results from this study can be used by policy makers, health planners, and insurers to plan and improve delivery of health services.
(b) Brain-based Biomarkers for Depression Diagnoses
We built predictive and descriptive models to diagnose depressed individuals based on the EEG signals recorded of brain activities in three frequency bands (Alpha, Beta, and Theta). This study identifies significant biomarkers in the EEG signals that can help physicians in their decision making. The data was analyzed through the use of various analytical models and data mining techniques including neural networks, decision trees, and k-means clustering. The findings of this study will contribute to a larger clinical trial, aiming to determine whether treatment with two antidepressants is more effective than treatment with only one.
Bijan Raahemi is an associate professor at the Telfer School of Management, University of Ottawa, Canada, with cross-appointment with the School of Electrical Engineering and Computer Science. He received his Ph.D. in Electrical and Computer Engineering from the University of Waterloo, Canada, in 1997. Prior to joining the University of Ottawa, Dr. Raahemi held several research positions in Telecommunications industry, including Nortel Networks and Alcatel-Lucent, focusing on Computer Networks Architectures and Services. His current research interests include Data Mining, Machine Learning, Information Systems, and Data Communication Networks. Dr. Raahemi has established the Knowledge Discovery and Data mining (KDD) lab at the University of Ottawa. His work has appeared in several peer-reviewed journals and conference proceedings. Dr. Raahemi also holds several patents in Data Communications. He is a senior Member of the Institute of Electrical and Electronics Engineering (IEEE), and a member of the Association for Computing Machinery (ACM).
Expanded Bio: http://web5.uottawa.ca/www5/braahemi/biography.htm