Stroke Outcome Prediction Using Vascular Imaging Features and Machine Learning

This technology is a stroke outcome prediction method to determine patient outcomes and inform treatment decisions. The method utilizes machine learning models trained with patient-specific vascular imaging features, patient demographics, stroke severity scores and clinical history as predictors to forecast a 90-day functional outcome following an ischemic stroke. 

Background: 
Strokes are the fifth leading cause of death in the United States and are also a leading cause of disability. Strokes occur when the arteries leading to and within the brain are blocked by a clot or bursts causing oxygen deprivation to the brain. Stroke prediction is difficult because they result from many different pathophysiologies. 

When caught early, strokes are preventable, and researchers have begun to tap into the machine learning world as well as artificial intelligence to look for these biomarkers and risk factors. By integrating machine learning models, physicians can evaluate the possibility of a future stroke with more patient-relevant information, such as patient history, that is important to determining the patient’s risk of stroke. From these models, there is an ability to incorporate them into apps or software for even easier processes. 

Applications: 

• Stroke outcome prediction 
• Post-stroke personalized treatment 
• Potential novel indications in other diseases (e.g., cardiovascular)

Advantages: 

• Personalized patient outcomes based on their history, images, and demographics
• Higher accuracy than current models
• Provides data relevant to treatment decisions
• Analyzes information from multiple sources in prediction model