Machine Learning-Based Risk Stratification for Symptomatic Intracranial Hemorrhage and 3-Month Prognosis following...

Area of Science:

• Neurology
• Medical Informatics
• Biostatistics

Background:

• Intravenous thrombolysis (IVT) with tissue-type plasminogen activator (tPA) is crucial for acute ischemic stroke but carries a risk of symptomatic intracranial hemorrhage (sICH).
• Predicting sICH and functional recovery (modified Rankin Scale - mRS) is vital for optimizing treatment and patient outcomes.
• Existing scoring tools have limitations in accurately predicting post-tPA complications and outcomes.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting sICH and 3-month functional outcomes after tPA.
• To identify key prognostic variables influencing sICH risk and long-term recovery.
• To compare the performance of developed models against established scoring tools.

Main Methods:

• Analysis of data from 434 ischemic stroke patients treated with tPA.
• Development and evaluation of three supervised classification models: Logistic Regression, Random Forest, and XGBoost.
• Five-fold cross-validation used to assess model performance (AUC, accuracy, recall, precision) and comparison with six existing scoring tools.

Main Results:

• Machine learning models (XGBoost AUC 0.89, Logistic Regression AUC 0.87, Random Forest AUC 0.82) outperformed conventional scoring tools in predicting sICH.
• The 24-hour National Institutes of Health Stroke Scale (NIHSS) score was the most significant predictor for both sICH and 3-month outcomes.
• Factors associated with increased sICH risk included prior stroke history and male sex; increasing age correlated with poorer 3-month outcomes.

Conclusions:

• Machine learning models demonstrate high accuracy in predicting sICH and 3-month outcomes post-tPA, offering superior performance to existing tools.
• The 24-hour NIHSS score is identified as a critical prognostic indicator.
• These models show potential as decision-support tools for personalized management strategies in acute ischemic stroke patients receiving thrombolysis.