Enhancing Stroke Risk Prediction: Leveraging Machine Learning and MRI Data for Advanced Assessment

As a leading cause of death across the globe, strokes have historically been regarded as a dangerously impactful condition with little to no predictability. Currently, the few ways that people can tell whether or not a stroke is taking place in a patient are through questionable methods, such as looking for warning signs and checking for hereditary factors. We aim to create a quantitative approach to this problem by offering a tool that neurologists will be able to take advantage of, providing them with the data they need to predict strokes before they happen. We intend to use MRI scan data obtained from OpenNeuro, specifically showing brains of adult patients pre-stroke and post-stroke. We will use this data to train a variety of machine learning models independently, including Support Vector Machines (SVM), K-Nearest Neighbors (KNN), and Random Forest. By training these models, we aim to accurately predict the risk of a stroke by inputting healthy brain imaging scans and determining their similarity to post-stroke brain scans. If implemented in neurologist offices, this model could have a considerable effect, allowing people to realize that a stroke is imminent and to begin preparing for it, potentially saving lives and improving outcomes.