Grant: Project Grant

When someone suffers a serious brain event like a brain hemorrhage, clinicians use a type of exam called a head CT scan to make the diagnosis and look for complications. Accurate prediction and identification of these complications, such as increased brain hemorrhage, are essential to limiting their impact. With the rapid progress in Artificial Intelligence (AI), such technology has the potential to assist physicians in these tasks and lead to improved diagnosis and treatment. In this project, we aim to train an AI model that can automatically characterize brain hemorrhage on CT scans and predict the risk of complications. More specifically, this requires the following steps: 1.Prepare a dataset: We will gather hundreds of CT scans and label the images so AI models can learn to predict and identify complications. This includes images from public sources and our hospital. 2.We will show the labelled dataset to the model so it can learn to analyze CT scans and predict complications. We will also use methods to help the AI model perform better in situations it didn't see often during training. 3.The AI model will be evaluated on images from sources other than the one it was trained on to verify whether it is robust enough to be used in different hospitals. 4.Finally, we will evaluate the performance using more recent data to ensure that it maintains its current level of performance over time. Our team includes experts in brain imaging, critical care medicine, neurology, and computer science. Together, we are working to connect AI applications with real-life patient care to improve the prediction and monitoring of complications in patients experiencing stroke. Following this project, we plan to use our AI models in different Canadian centers to evaluate the impact on diagnosis and clinical decisions.