AI could help doctors make better healthcare decisions and treat patients proactively

hospital

While artificial intelligence gets a lot of attention for its capacity to transform wide-ranging industries, it’s difficult to predict AI’s true impact. As Integrate.AI’s Kathryn Hume puts it, the term itself is murky enough that we can exercise agency in how we choose to define it. We can also choose how this technology can and should be applied.

Several weeks ago at SickKids hospital in Toronto, several researchers, doctors, and physicians convened to discuss AI’s potential impact on the healthcare industry for the hospital’s weekly Grand Rounds series, which puts the spotlight on new discoveries and advances in child health from researchers around the world.

“At the end of the day, we think about adaptive decision management to think about tackling those unpredictable things in a more predictable way.”
– Lianne McLean

SickKids, which is the subject of the Toronto tech community’s $25 million charity initiative, is itself working to bring its institution to the 21st century; the 68-year-old hospital wants to use the funding of #Tech4SickKids to create a chair in bioinformatics and AI; support the hospital’s big data and analytics initiatives; and construct a state-of-the art emergency suite.

In a room packed with SickKids doctors and nurses, the speakers provided a glimpse into the landscape of AI and healthcare research in Canada, and how the technology could make the jobs of doctors and nurses much easier.

To separate the science from the science fiction, SickKids critical care physician Mjaye Mazwi gave a rundown on the basics of how AI works, and how the technology could augment the job of a doctor. The best use cases, according to Mazwi, is the ability to assist in the decision-making of treating patients; the decision-making process of an AI is not unlike that of a physician.

“When you perform an exam and try to elicit clinical signs and ask your patients about their symptoms, you’re extracting features, which you then pass through a classifier that you developed in medical school to try to decide whether a disease entity is present,” said Mazwi.

“Our goal is to think about if we can use machine learning to aid in our decision making.”
 

Helping doctors act proactively instead of reactively can also help patients have better experiences in the hospital. Lianne McLean, an emergency physician at SickKids, relayed the common experience of the emergency room. Generally, the hospital knows that winter will be busy, so there will be more staff on site in preparation for the season. However, there are still some instances where they don’t have foresight.

“Our goal is to think about if we can use machine learning to aid in our decision making, and we say aid because we do use all of these things that are tangible in our brains to make decisions about staff distribution and where patients are going to go. But at the end of the day, we do think we can do better,” she said.

McLean’s team is currently working with a US-based startup called Qventus, which can take historical data from the last 100 days and anticipate patient needs in the future.

“At the end of the day, we think about adaptive decision management to think about tackling those unpredictable things in a more predictable way. We live in a space where we don’t have control of a lot of our space, but how can we best equip ourselves to make better decisions?”

SickKids is already working to be a part of research in AI and healthtech. Anna Goldenberg, a University of Toronto assistant professor of computer science and scientist in the Genetics and Genome Biology Lab at SickKids Research Institute, is also working with Toronto’s Vector Institute. She’s working on creating machine learning methods that make such data useful for clinical diagnosis, and talked about the work she’s doing at the Institute.

Of course, when it comes to having access to a wide breadth of data, privacy is always a concern. Goldenberg, who said that having streamlined access to data is currently a challenge, said that it’s important to build ethical processes, but that releasing data for researchers is also critical.

“I think there is an importance of doing it right and following specific guidelines. These things change slowly but I feel like the process is started, and there’s a lot of inertia and momentum in the field right now,” she said. “Several hospitals have had some successes integrating machine learning into their practice, so people can see that it’s beneficial to the patient and to the doctor, and ethics will make sure that the guidelines are such that it only benefits. Ss much as possible, the risks are mitigated, but it will have to happen.”

This photo was originally published on Unsplash