In the latest episode of Approximately Correct, Amii Fellow and Canada CIFAR AI Chair Russ Greiner talks about how machine learning can transform how we view medical diagnosis and treatment. 

For years, Greiner has been captivated by the potential of machine learning to improve survival predictions, developing tools that provide patients and medical professionals with personalized health forecasts. While risk scores, predictions, and statistical models have always been integral to medicine, they often rely on generalized data. For example, a patient diagnosed with a new condition might receive information on median survival times, derived from the outcomes of other patients with similar diagnoses and treatments.

But Greiner says those are very broad estimates that might not reflect a particular person’s situation. 

“So you get a survival curve for, say, stage three breast cancer that says … if everyone was identical, this is what you'd expect. We're not all identical. There are different people, different characteristics and some will die earlier, he says.

“Can we take that idea and say I can predict for Mr. Smith, his curve, looks like that. And Miss Jones her curve looks like that.”

Revolutionizing Survival Prediction

Greiner’s research focuses on work that leads to actionable insights and AI models that answer critical questions: What treatment should a patient undergo, and what are the expected outcomes?

By leveraging survival prediction models, Greiner aims to empower patients and doctors to make informed decisions tailored to individual needs. And he’s found that once medical providers see the benefits that machine learning and survival prediction can offer, they are enthusiastic.

“I’ve worked with a lot of doctors. Once I describe the work, they say ‘Give me one of those,” he says.

To learn more about Greiner’s work on survival prediction, check out this primer on the topic, or his presentation on Learning Models that Predict Objective, Actionable Labels.

Approximately Correct: An AI Podcast from Amii is hosted by Alona Fyshe and Scott Lilwall. It is produced by Lynda Vang, with video production by Chris Onciul.

Listen and subscribe to Approximately Correct on Spotify, Apple Podcasts, or your favourite podcast platform.