Recent advances on ischemic stroke research highlights that utilizing a careful treatment selection routine influences the patient outcome. Treatment selection can be improved using the information available in the CTA images, but optimally making use of the representative information in the images requires developing novel imaging biomarkers. Automated approaches using these features are going to enable the development of powerful prediction models, and augment the decision making step in the clinical practice.
The aim of the project is to first develop and experimentally verify a set of approaches for the automated analysis of the CTA data acquired in the acute stroke phase. Our analysis includes exploration of collateral vessel scoring methods which represent the flow in the collateral regions. Being computationally compact, our biomarkers are going to be the input for training prototype convolutional deep learning models specialized to predict the outcome status. Final phase of the project focuses on the assessment, and quantifies the contribution of novel biomarkers.