Utilizing imaging parameters for functional outcome prediction in acute ischemic stroke: A machine learning study.

Area of Science:

• Neuroimaging
• Machine Learning in Medicine
• Stroke Research

Background:

• Acute ischemic stroke with anterior circulation large vessel occlusions (LVOs) presents a significant clinical challenge.
• Predicting functional outcomes is crucial for guiding treatment decisions.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting functional outcomes in acute ischemic stroke patients with LVOs.
• To assess the predictive value of imaging parameters alone, irrespective of treatment or stroke severity.

Main Methods:

• A retrospective study of 180 adult patients with anterior circulation LVOs who underwent CT angiography (CTA) and CT perfusion.
• Machine learning algorithms (CatBoost, XGBoost, Random Forest) were trained using only imaging parameters.
• Model performance was evaluated using AUROC, AUPRC, accuracy, Brier score, recall, and precision, with SHapley Additive exPlanations for feature importance.

Main Results:

• The XGBoost model achieved the highest predictive performance with an AUROC of 0.91.
• Key performance metrics included precision (0.72), recall (0.81), AUPRC (0.83), and accuracy (0.78).
• Multiphase CTA collateral score emerged as the most significant imaging feature for outcome prediction.

Conclusions:

• Machine learning models utilizing solely imaging parameters can effectively predict functional outcomes in acute ischemic stroke with LVOs.
• The high predictive accuracy (AUROC 0.91) suggests imaging parameters are comparable to conventional predictors.
• The multiphase CTA collateral score is a critical imaging biomarker for predicting stroke recovery.