five ways artificial intelligence is improving health care in canada
hospitals are starting to implement artificial intelligence (ai) to help save lives – one of many advances in the controversial technology that entered our consciousness in 2023.
5 minute read
there are wearables that track your heart rate, robots that perform surgeries and tools that predict the next respiratory outbreak – all adding to the tremendous potential of transforming canada’s ailing heathcare system.
getty images
hospitals are starting to implement artificial intelligence (ai) to help save lives – one of many advances in the controversial technology that entered our consciousness in 2023. the year began with many befriending chatgpt, the online ai-powered chatbot developed by openai. as 2023 comes to a close, ai has infiltrated every aspect of our daily lives, for better or worse.
there are wearables that track your heart rate, robots that perform surgeries and tools that predict the next respiratory outbreak – all adding to the tremendous potential of transforming canada’s ailing heathcare system.
looking back at 2023, and forward to 2024, canadian researchers share the innovative ways ai can be used to improve health for canadians.
dr. ross mitchell; university of alberta, alberta health services
a couple of years ago ross mitchell returned to canada to advance ai in health with alberta machine intelligence institute (amii) – one of the three leading ai centres in canada.
mitchell was particularly attracted to alberta’s data warehouse, which holds health information collected over the last 20 years from 4.7 million albertans.
applying techniques to combine patient records, medications, and scans, helps doctors better understand patient risks, and diagnose and treat patients more quickly.
advertisement
for example, when colleagues created a registry to study inflammatory bowel disease (ibd) in 2021, they had a repository of 20 years of data from 92,000 ibd patients plus 160,000 abdominal scans from 59,000 of those patients.
for comparison, the uk biobank project received the equivalent of $50 million (cad) to acquire an additional 60,000 abdominal ct scans.
“it’ll probably take them about 5 years to collect the data,” says mitchell, something alberta health services has already accomplished.
elham dolatabadi; york university, vector institute
elham dolatabadi, a researcher at york university and vector institute, highlights innovative ai solutions for mental health. dolatabadi had already been involved in developing conversational ai agents for customer services using large language models. kids help phone reached out to see if these techniques could help frontline staff.
“think of it as a second pair of eyes. like someone sitting beside you shares their thoughts in real time. and can pull up information for you,” she says.
it would make staff more user-centered: the ai agent organizes the conversation into issues to be discussed and prompts staff with personalized resources.
advertisement
for example, if a caller speaks about depression, the ai agent can identify relevant resources, picking up keywords.
currently, the team is testing the agent by comparing the reactions to calls with staff reactions. it also helps the researchers learn more about how ai agents understand context in conversations and find ways to make agents fair and responsive.
in 2024, dolatabadi will be focusing on three ds – deployment, democratizing and demystifying ai.
dr. devin singh; the hospital for sick children (sickkids)
as a resident, one of his patients died while waiting for care. this lit a fire in the er doctor. now at toronto sickkids, dr. singh is focused on reducing deaths due to wait times, backlogs and inefficiencies. he has started hero ai, a company that helps hospitals use ai.
the standard er practice goes like this: “you go through triage, and then you’re stuck waiting .. and then you see a doctor, or a resident. they say, ‘oh you might have appendicitis. let’s order an abdominal ultrasound’,” singh says. by then it might be hours later.
“our proposal is to say, ‘let’s just order the test in an automated way’. get your ultrasound while you’re waiting, and when you see the physician, you have your testing done and a diagnosis already.”
advertisement
changing the hospital workflow expedites the next steps: surgery, consultations and getting antibiotics, if necessary.
timothy chan; university of toronto, data sciences institute (dsi), centre for analytics and artificial intelligence engineering (carte)
throughout his career, timothy chan has been using ai to solve health-care problems. the professor at the university of toronto jumped at the chance to help milk banks using ai.
“it’s something close to my heart as well, because my son was born prematurely and spent several weeks in the nicu,” says chan.
oftentimes, premature babies in neonatal intensive care units (nicu) don’t get their mother’s milk. the roger hixon ontario human milk bank at mount sinai hospital opened in 2013, and distributes breast milk to 50 nicus, reducing medical complications for the tiniest patients.
macronutrients like fats and protein are critical for preemies. however, donor information is general. identifying nutrients in donations ensures uniform distribution of protein in the batch after blending.
“donations from several mothers makes a more uniform product because there’s quite a bit of variability in the nutrients and other things in each individual donation, from mom to mom, even within a mom, from day to day,” chan says.
advertisement
analyzers to assess nutrition in milk donations are expensive, and don’t provide staff with the recipe for blending donations into a batch. chan’s team used ai to predict content in the milk using demographic information from donors. a trial was carried out for a year and half, showing batches made using ai made blending faster and improved the nutritional content of the milk.
in the new year, chan says the ai tool will be implemented at roger hixon. he hopes to use ai to help processes such as predicting bacteria in the milk.
dr. jacqueline kueper; western university, institute for better health
for six years jacqueline kueper has been working with the alliance for healthier communities , an organization supporting ontario community health centres (chc).
“initially we actually failed a lot [when developing ai models],” she says. with 11 years of data from 72 chc, ai techniques could help them understand the care setting and its complexity better than traditional methods.
the current practice-based learning network (pbln) team narrowed the initial focus to predicting mental health needs for people with diabetes. but they worried that there wouldn’t be capacity to serve all the extra clients that would get flagged. the team refined the problem to target population-level planning and advocacy.
advertisement
“i am not coming in as an ai researcher, with access to data and just developing a tool to say, ‘look this is going to be great for you’. it’s saying, let’s engage with the community and the health system and the provider there and figure out what are actually meaningful things to try and target,” kueper says about the iterative process.
ayeshah haque is a fellow in health journalism at the university of toronto, and the former community engagement coordinator for the ai4ph national training program.
share story
