Early Prediction of Cerebral Vasospasm After Aneurysmal Subarachnoid Hemorrhage Using a Machine Learning Model and...

Area of Science:

• Neurology
• Neurosurgery
• Medical Informatics

Background:

• Cerebral vasospasm is a significant complication following aneurysmal subarachnoid hemorrhage (aSAH), leading to delayed cerebral ischemia (DCI) and poor neurological outcomes.
• Early prediction of vasospasm is challenging due to its complex etiology but is critical for effective patient management and timely therapeutic interventions.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for the early prediction of cerebral vasospasm in patients with aSAH.
• To identify key clinical and imaging predictors of vasospasm using routinely collected patient data.

Main Methods:

• Retrospective analysis of 345 patients with angiographically confirmed aSAH.
• Extraction of demographic, clinical (Hunt and Hess grade, Fisher scale), and treatment (external ventricular drainage - EVD) parameters.
• Training and evaluation of seven supervised ML models, including logistic regression and gradient-boosted trees, using nested cross-validation.

Main Results:

• Over 50% of aSAH patients experienced moderate to severe vasospasm.
• Independent predictors of vasospasm included younger age, higher Hunt and Hess and Fisher grades, and EVD placement (p < 0.001).
• Logistic regression demonstrated the best discrimination (AUC-ROC 0.723), while tree-based models offered higher sensitivity (0.867). Aneurysmal etiology increased vasospasm risk (OR 4.72).

Conclusions:

• Routinely available clinical and imaging data are sufficient for reliable ML-based vasospasm prediction post-aSAH.
• Logistic regression offers a balance of accuracy and interpretability, whereas tree-based models prioritize sensitivity for vasospasm prediction.
• An interpretable ML tool utilizing routine data could support bedside vasospasm prediction, warranting prospective validation.