Grant: Project Grant

Every five minutes in Canada, someone has a stroke. Ischemic stroke, the most common stroke, happens when a blood clot blocks a brain vessel. Therapies like clot-busting drugs and thrombectomy can help, but carry risks like brain bleeding. Still, 1 in 2 patients lose their independence and 1 in 5 die within 90 days. Thus, when a stroke happens, patients and families want personalized answers to 3 big questions to guide decisions: 1) Will this person get better with treatment? 2) Will they be harmed (will they bleed?) 3) Will they die, or perhaps worse, survive with extreme disability? Brain scans can show how much brain is affected by stroke, prior injuries, and poor blood flow. These imaging features, along with aspects of the patient's history and examination, influence recovery, bleeding risk, and survival. However, accurately measuring these features is difficult and time consuming, even for experts. As a result, important information from brain scans is not used, and doctors rely on crude outcome prediction tools that are often inaccurate. The ACT-RECOVER study aims to change this by using artificial intelligence (AI) to combine the most useful imaging and clinical data to rapidly and accurately predict recovery, brain bleeding, and death or severe disability after stroke. Our goal is to provide personalized outcome predictions that guide patient care. We will use high quality data from prior stroke trials to build AI models that predict 90 day outcomes. We have built tools to automatically identify key stroke features on brain scans. We will use them to provide imaging information to our prediction models. We will then test the predictions in an international trial of 5000 patients. ACT-RECOVER will help patients and families have clearer conversations about outcomes and enable better shared decisions about stroke care. Our work will also make stroke trials more efficient and safer by allowing them to predict their results and identify safety concerns early.