In order to fully capitalize on the power of predictive tools such as Artificial Neural Networks, an expert system is need to process the outcome information and interpret to answer a specific medical decision. The medical decision which I am working on is the ethical dilemma of whether to initiate/continue or to terminate aggressive treatment for preterm babies. Our goal is to develop a generic tool that incorporates our research group's existing predictive tools to help physicians and parents cope with this ethical decision. This tool will be adapted to various situation and ensure that parents are included in the decision-making process and are able to reach a consensus with the physcians, without coerion.

Lan Yang, Dr. Monique Frize

Positron Emission Tomography (PET) is currently being used at the Ottawa Heart Institute for quantification of myocardial perfusion - to measure blood flow through the heart. The radioisotope that is being used (for a variety of good reasons) is rubidium-82. There is an inherent problem with cardiac PET images based on this technique, however - the images produced are very noisy. During the last 10 years, image noise reduction (known more commonly in the literature as denoising) based on the discrete wavelet transform (DWT) has shown to be a promising technique. This is because of a the wavelet representation's ability to better separate noise energy from signal energy. The focus of my thesis project will be to research the most recent DWT-based denoising strategies and determine their potential applicability to these cardiac PET images. Better images will result in more accurate blood flow estimates, which increase the physician's ability to recommend the appropriate course of treatment.

Geoff Green, Dr. Aysegul Cuhadar

Medical thermography is a non-invasive, non-radiative imaging procedure that records the thermal distribution of human body by use of an infrared camera. The thermal distribution of a person is closely related to physiological processes and can provide valuable information about the initiation and progression of functional abnormalities or pathologies.

My research deals with the automated interpretation of medical infrared images. I am interested in all the steps of the image processing chain, from the preprocessing to the final suggested diagnosis. The goal is to reduce the errors introduced by current manual assessment, and to save valuable time to health-care specialists. Current applications of medical thermography investigated are the assessment of pain, breast cancer, rheumatoid arthritis. The assessment of pain is of particular interest since the presence and location of pain is a difficult and highly subjective task in current health care practice.

My research interests also include artificial intelligence and digital signal processing applied to medicine.

Christophe Herry, Dr. Monique Frize, Dr. Rafik Goubran

An electronic nose (Enose) is "an instrument which combines an array of electronic chemical sensors with partial specificity and an appropriate pattern-recognition system, capable of recognizing simple or complex odors". It has been widely used in food processing, environmental monitoring, fragrance development and also medical diagnostics. My research is using a sensor array of 32 smell-sensitive sensors (manufactured by Cyranose company, USA), developing and optimizing the signal processing procedure to increase the classification accuracy. Currently implemented method include DSP, statistics method and Neural network tools, future consideration including Fuzzy logic and Genetic algorithm. Current experiment is still on lotions with different smell, actual clinical samples in consideration include smell difference between healthy people and heart-failure patient, or even between heart-failure patients in different stage.

Hong Chen, Dr. Rafik Goubran, Dr Toffy Mussivand in Ottawa Heart Institute