Predictive Models for Secondary Epilepsy in Patients with Acute Ischemic Stroke Within One Year
Abstract
Objective
Post-stroke epilepsy (PSE) is a major complication that worsens both prognosis and quality of life in patients with ischemic stroke. This study aims to develop an interpretable machine learning model to predict PSE using medical records from four hospitals in Chongqing.
Methods
We collected and analyzed medical records, imaging reports, and laboratory test results from 21,459 patients diagnosed with ischemic stroke. Traditional univariable and multivariable statistical analyses were performed to identify key predictive factors. The dataset was divided into a 70% training set and a 30% testing set. To address class imbalance, the Synthetic Minority Oversampling Technique combined with Edited Nearest Neighbors was used. Nine widely applied machine learning algorithms were evaluated and compared using relevant prediction metrics. SHAP (SHapley Additive exPlanations) was used to interpret the model, assessing the contributions of different features.
Results
Regression analyses showed that complications such as hydrocephalus, cerebral hernia, and deep vein thrombosis, as well as brain regions (frontal, parietal, and temporal lobes), significantly contributed to PSE. Factors like age, gender, NIH Stroke Scale (NIHSS) scores, and laboratory results such as WBC count and D-dimer levels were associated with a higher risk of PSE. Among the machine learning models, tree-based methods such as Random Forest, XGBoost, and LightGBM demonstrated strong predictive performance, achieving an AUC of 0.99.
Conclusion
Our model successfully predicts PSE risk, with tree-based models showing superior performance. The NIHSS score, WBC count, and D-dimer were identified as the most important predictors.
Related articles
Related articles are currently not available for this article.