Machine learning-based prediction model for post-stroke cerebral-cardiac syndrome: a risk stratification study.

Area of Science:

• Neurology
• Cardiology
• Artificial Intelligence

Background:

• Cerebral-cardiac syndrome (CCS) is a serious complication of ischemic stroke.
• Early prediction of CCS is crucial for managing adverse outcomes.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting CCS in ischemic stroke patients.
• To identify key predictors of CCS using ML interpretability techniques.

Main Methods:

• Retrospective analysis of 511 ischemic stroke patients.
• Development and comparison of five ML models (Logistic Regression, Random Forest, SVM, XGBoost, DNN).
• Evaluation of model performance using AUC, accuracy, precision, recall, F1 score, and calibration metrics (Hosmer-Lemeshow, Brier score).

Main Results:

• XGBoost demonstrated the highest discrimination (AUC 0.879), followed closely by Random Forest (AUC 0.866).
• SVM and Random Forest showed optimal calibration.
• Key predictors identified include D-dimer, ACEI/ARB use, HbA1c, C-reactive protein, and prothrombin time.

Conclusions:

• ML models accurately predict early CCS in ischemic stroke patients.
• XGBoost offers superior predictive discrimination, while SVM and Random Forest provide better calibration.
• Integrating these ML models can improve risk stratification and guide preventive strategies for CCS.