Automatic collateral circulation scoring in ischemic stroke using 4D CT angiography with low-rank and sparse matrix...

Area of Science:

• Medical Imaging
• Neurology
• Artificial Intelligence

Background:

• Effective endovascular treatment for ischemic stroke relies on adequate collateral blood supply.
• Current collateral scoring methods based on visual inspection exhibit inter- and intra-rater variability.
• Objective and consistent collateral assessment is needed for optimal stroke treatment decisions.

Purpose of the Study:

• To develop and validate a robust, automated method for scoring cerebral collateral blood supply.
• To improve the accuracy and consistency of collateral assessment in ischemic stroke patients.
• To aid clinical decision-making in endovascular stroke treatment.

Main Methods:

• The Automatic Collateral Circulation Evaluation in iSchemic Stroke (ACCESS) method utilizes 4D dynamic CT angiography (CTA).
• It employs a fast robust matrix completion algorithm with Frank-Wolfe optimization to model unfilled and full cerebrovasculature.
• The collateral score is calculated as the ratio of unfilled to full vasculature, mirroring clinical protocols.

Main Results:

• ACCESS achieved an overall accuracy of 84.78% in a cohort of 46 stroke patients.
• Leave-one-out cross-validation yielded a mean area under the receiver operating characteristics curve of 85.39%.
• The method demonstrates high performance in differentiating collateral supply levels.

Conclusions:

• ACCESS provides an automated and objective approach to collateral scoring, overcoming limitations of manual assessment.
• The method's robustness and resemblance to radiologist scoring facilitate its integration into clinical workflows.
• This tool can assist radiologists in making more informed treatment decisions for ischemic stroke.