Skip to main content Skip to Footer

BLOG


November 27, 2018
AI in Healthcare: The challenge of improving lives without compromising individual privacy
By: Jodie Wallis

The AI Effect is a podcast series exploring Canada’s burgeoning artificial intelligence ecosystem. Accenture’s AI leader co-hosts with reporter Amber Mac to look at AI’s explosive growth and the change—challenges and rewards—it can bring for individuals, business and society.

In Episode 4 we spoke with experts about how AI is being adopted in healthcare … and making a difference in people’s lives.

Some of most important changes coming to our lives through AI are in healthcare—from simple visits to the doctor to the most complex diagnoses and even predicting and preventing diseases. Anna Goldenberg, senior scientist at SickKids Research Institute in Toronto, is using AI in health and biology to detect patterns in complex data that the human mind cannot readily find.

Goldenberg and her team, for example, are using AI to predict whether a child who has just broken their leg in a ski accident is at risk of cardiac arrest five minutes before it might happen—and be right 70 per cent of the time.

They are also working on using AI to ease the burden of certain tests, like MRIs in really small children for example. Using blood, the machines can detect how likely a child with genetic predisposition is to get cancer before the age of 6. If likely, they can be referred to the standard protocol. If not, they can be spared the pain of more complex tests until they are old enough to handle them better.

The predictive abilities of AI can also be used to look for patterns of human behaviour. Kenton White, co-founder and chief scientist at Advanced Symbolics Inc., explained to us how a pilot project is analyzing peoples’ blog posts and social media activity to detect and classify depression, PTSD, insomnia and other factors that can lead to suicide. The data can’t predict individual risk, but it is helping to identify where to concentrate resources for suicide prevention at the community and cluster level.

When talking about healthcare, it’s almost impossible not to bring up the question of privacy. Our health data is some of the most private data there is about ourselves. But, are we at risk of letting privacy concerns get in the way of progress? White explains that by looking at populations in the aggregate to model disease and behaviour, we can put safeguards in place like minimum population size and double-blind sampling that will get us to positive outcomes without compromising individuals’ privacy.

And there are other deep ethical issues to consider, especially interpretability and accountability. Goldenberg lays out the question like this: “How do we know that they are basing their decisions on something reasonable, especially when it’s not how a human would have made the decision? And who takes responsibility if the machine is wrong if we don't understand it?” The predictive algorithms being developed can be so complex that, even if the analysis is right, it can be difficult for humans to interpret what the machines have found.

Nevertheless, AI is already making a positive difference in healthcare, and it’s bound to make larger and faster inroads to improving peoples’ lives, in emergencies and day-to-day.


Popular Tags

    More blogs on this topic

      Archive