Automated analysis tools for assessing the vasculature in imaging for acute stroke

The project concerns the development and evaluation of image analysis tools for studying the vasculature in cases of acute ischaemic stroke. A starting point would be the scoring of collateral flow in CT angiography (CTA) datasets. In acute ischaemic stroke, collateral flow via leptomeningeal vessels maintains viable brain tissue (the “ischaemic penumbra”) for a variable period of time. Treatment responses to reperfusion are better in those with good compared to poor collaterals, so collateral flow assessment may therefore be critical in the acute decision making stage after stroke.

A method of scoring has been developed at the University of Glasgow, but this involves many manual steps. To fully automate these steps, the cerebral vasculature and circulation territories need to be segmented, occlusion sites identified and careful contra-lateral analysis performed. This represents a significant image analysis challenge, possibly requiring atlas based and machine learning approaches.

Once a software tool has been developed on available acute stroke data, the automated scoring system will require testing against expert readers in observer agreement studies that can utilise CTA from the acute stroke database at the University of Glasgow. Comparison against data from acute angiograms in the PISTE multicentre trial will provide additional validation against an expert panel of readers and conventional angiography. Validation against clinical outcomes and other imaging variables, including CT perfusion, will be obtained from the South Glasgow Stroke Imaging database.

When applying, please search using "automated analysis" and select the full project title as it will appear below.