Dr. Bo Wang, UHN's chief AI scientist.Tim Fraser/Supplied
For Geoffrey Hinton, improvements to human health top the list of benefits that will stem from the growing use of machine-learning algorithms.
It’s a point the artificial-intelligence trailblazer emphasized on Tuesday when he was dialled into the press conference that announced him as a co-recipient of this year’s Nobel Prize in physics.
In practice, AI is making its presence known everywhere in our health system, but its impact is just beginning to be felt. To get a snapshot of the transformation, The Globe and Mail spoke with Bo Wang, chief artificial-intelligence scientist at University Health Network in Toronto, the country’s largest research hospital.
Geoffrey Hinton, who warned of AI’s dangers, co-wins Nobel Prize in Physics
Artificial intelligence is already hard at work in many sectors – for good and bad
Last year you became the first chief AI scientist at a Canadian hospital. How would you describe the job?
It’s definitely a very exciting role for me. One of the main aspects of the job is to lead the integration of AI into a clinical setting. We aim to improve the accuracy, efficiency and personalization of patient care at all the hospitals across UHN. I am also a scientist, so I work at the forefront of developing and implementing advanced AI models that are particularly tailored to the health care sector.
Technical advances are a part of modern medicine. Is there something about AI that’s different from other breakthroughs when it comes to health care?
One difference is that AI is very data-driven. There’s a famous saying in computer science: garbage in, garbage out. That’s why we need to make sure AI models are trained using proper data and validated. A second difference is that AI is capable of finding subtle patterns in data that are hard for humans to find – that really sets it apart from traditional techniques based on statistics.
Where do you see AI having the biggest impact in hospitals?
Broadly speaking, AI is making strides in a few areas. The first is imaging. It was predicted a few years ago that AI will greatly revolutionize radiology. I think this is pretty much true, in the sense that AI is already widely used to analyze medical images such as X-rays, MRI and CT scans to assist radiologists diagnosing diseases, like cancer, and to reduce diagnostic errors. The second is natural language processing, which can be employed to extract meaningful insights from clinical notes, assisting with documentation and predicting patient outcomes. And a third area is personalized medicine, where AI is used to analyze genetic information and other data to guide treatment plans for individual patients, particularly in oncology.
To what extent is AI having an impact on patients right now?
I think it’s still in an early phase. There are some innovations we’ve made at UHN that utilize AI to improve patient care here. One example I can give is the analysis of blood test results and daily clinical variables collected from cardiac patients. This has been widely used across hospitals in Canada and is being used by nurses and doctors to better manage patients with heart failure. Another area is AI for surgery. We see robotics and AI continuing to improve surgical procedures to reduce recovery times and improve patient outcomes.
Does that mean using AI to plan a surgery or during surgery?
It’s actually both. We have projects that help with surgery planning based on 3-D scans of patients. And once they are prepared, we also have other AI models that help surgeons with real-time decision making.
How does the emergence of AI in the hospital setting affect the doctor-patient relationship?
As we know, in Canada we face a health care crisis. We have lots of burnout across medical professionals. One of the major factors driving doctor burnout is charting. Basically, doctors spend 15 to 20 minutes per patient visit typing, to keep notes. Nowadays, AI can be used to automate these mundane tasks to help them do a better, faster job at these administrative tasks so that they have more time for the most important thing they sign up for, which is patient care.
How should we separate what we hear about the promise of AI in health from the hype?
There’s definitely a certain level of hype. That’s always been there. But I do see practical applications of AI in day-to-day patient care. So in that sense, it’s already making a difference. Certainly there’s lots of potential risks, including data privacy, bias, lack of interpretability, or also lack of proper oversight. What I would say is we should work on both sides. One is the development of more accurate, more advanced AI tools. Two is to develop guardrails around these models, around these applications to make sure they are properly regulated and properly deployed.
This interview has been edited and condensed.