Natural Sciences and Engineering Research Council of Canada
Symbol of the Government of Canada

Common menu bar links

Past Winner
2018 NSERC E.W.R. Steacie Memorial Fellowships

E.W.R. Steacie Memorial Fellowship

E.W.R. Steacie Memorial Fellowship

School of Computer Science

McGill University

When it comes to developing artificial intelligence, a central problem is teaching a machine how to make decisions and learn from the outcomes for the next time it encounters a problem. Robots and computer programs that observe and adapt to situations have the potential to transform many industries, perhaps none more than health care. Check-ups, diagnoses, even basic treatments could be administered by intelligent machines quickly and confidently, freeing up waiting rooms and hospital beds, and saving patients unnecessary trips to the doctor.

Joelle Pineau, a researcher at McGill University, is developing algorithms that enable machines to learn, plan, solve and decide in uncertain situations.

Pineau uses a mathematical framework called reinforcement learning (RL), in which an artificial intelligence (AI) agent can explore and interact with its environment, receiving a positive signal when it performs a good action. This encourages the agent to choose positive actions more frequently and become more adept at its functions as it learns. Pineau is applying RL to challenges in health care to develop innovative strategies for personalized treatments. She and her team have already developed a real-time adaptation of neurostimulation to treat epilepsy. They also analyzed data from clinical studies to evaluate strategies for drug treatment for depression and schizophrenia.

Now, Pineau and her team are turning their focus to developing personalized medicine. Using the huge amounts of data available about cancer and heart disease, they are developing a computational RL framework, combined with deep learning methods, that will extract information from a variety of data, including medical charts, X-ray images, lab reports and clinical notes, and discover new, effective treatment strategies. The goal is for an AI agent to be able to capture a patient's state of health, and be able to plot a course of action to treat and assist. This has the potential to profoundly impact how medical treatments are discovered and deployed, and pushes us closer to a higher quality of health care tailored to the individual.

Joelle Pineau
For additional information or to request photos of NSERC Prize Winners, please contact Media Relations.