Stroke is the second leading cause of death and the third leading cause of disability worldwide. Similar trends exist in the U.S., where it is estimated that every 40 seconds a stroke occurs, leading to a stroke-related death every 4 minutes. Afflicting approximately 800,000 Americans each year, time sensitive and widely practicable diagnostic algorithms are paramount. Despite increasing emphasis on the significance of blood flow characteristics in pursuing strokes, estimating regional cerebral blood flow (rCBF) inside the skull is no trivial task. Imaging techniques that provide rCBF distribution may be unavailable due to time constraint or access to facilities, or simply the expenses attached to it. In our preliminary studies, we have developed a model that can give blood flow estimates in real-time for the entire brain vasculature. Here, we propose a novel hybrid computational model that utilizes high-performance computing and machine learning to simulate the entire brain vasculature, as well as provide real time measures of regional CBF. We will use novel mathematical methods to determine the vascular geometry of each subject using our existing imaging database that includes CTA and CBF images from stroke patients, and will subsequently validate our simulations against the same dataset. Finally, we will use machine learning algorithms to extract patterns in stroke patients' blood flow and give estimates for stroke severity predictors on patient- specific measurements.
The objective of this project is to estimate blood flow properties in the brain and its correlation with stroke. We will simulate and validate a model to give similar information as the expensive perfusion scanning measurements. Through Machine Learning algorithms, we will provide severity estimates for stroke patients for more informed clinical decisions and better patient care.
