An exploration on the machine-learning-based stroke prediction model.

Area of Science:

• Biomedical Informatics
• Artificial Intelligence in Medicine
• Clinical Diagnostics

Background:

• Stroke is a leading cause of death and disability worldwide.
• Cytokine levels have emerged as potential biomarkers for various diseases, including stroke.
• Machine learning (ML) offers powerful tools for analyzing complex biological data.

Purpose of the Study:

• To develop and evaluate ML models for stroke prediction using cytokine profiles.
• To identify key cytokine biomarkers associated with stroke.
• To enhance clinical decision-making in stroke diagnosis and management.

Main Methods:

• Recruited 2346 stroke patients and 2128 healthy controls.
• Utilized clinical laboratory tests and demographic data for model development.
• Employed Random Forest (RF), Gradient Boosting Machine (GBM), and Support Vector Machine (SVM) algorithms.
• Evaluated models using ROC curves, AUC values, and calibration curves.

Main Results:

• RF model demonstrated superior performance over GBM and SVM in the training set (AUC, sensitivity).
• Key cytokine features identified: IL-6, IL-5, IL-10, and IL-2.
• The developed stroke prediction model showed good generalization ability on the test set.

Conclusions:

• ML models based on cytokine features are effective for stroke prediction.
• Cytokine profiles, particularly IL-6, IL-5, IL-10, and IL-2, hold significant potential as biomarkers for stroke.
• The validated models can aid clinicians in stroke diagnosis and risk assessment.