A machine-learning approach to personalized risk prediction for recurrent ischemic stroke using population-level event data and routine brain imaging
Ischemic stroke, caused by a blood clot in the brain, is a leading cause of death, disability, and dementia. One in four of these strokes happens in patients who already had a prior stroke or mini-stroke. These preventable “recurrent strokes” carry a 20% risk of death. Causes include blockages of neck or brain vessels, disease of small vessels of the brain, or atrial fibrillation (irregular heart beats). Routine brain imaging done for patients with stroke allows doctors to identify features of most of these causes. But it is very difficult for us to predict the risk of recurrent stroke for individual patients. It is also difficult to predict how specific treatments will reduce this risk – especially when patients have many possible stroke causes. High-quality data specific to women and non-White populations are also lacking. These issues limit our ability to advise patients about their risk and personalize their stroke prevention.
Using health records, we will find patients who had a stroke or mini-stroke in Alberta from 2017-2018. We will study the brain imaging and other tests that they had to investigate their stroke. We will analyze these tests in detail using novel automated tools to identify different potential causes of stroke. We will track these patients until December 2023 (5 years) to see how many had more strokes. We will use this large, detailed dataset to develop a tool to predict the risk of recurrent stroke in a personalized way. We will also be able to predict how a patient’s risk may change with specific treatments like surgery for a blocked carotid artery (neck vessel). This project will provide up-to-date data for more personalized prevention of recurrent strokes. It will also strengthen Canada’s position as a leader in imaging-guided stroke prevention.
Aravind Ganesh , University of Calgary
Partners and Donors