Development and Validation of Multiple Machine Learning Models Integrating Neutrophil-Lymphocyte Ratio for Prediction of Hemorrhagic Transformation After Intravenous Thrombolysis in Acute Ischemic Stroke.

backgroundHemorrhagic transformation (HT) is a critical complication of intravenous thrombolysis (IVT) in acute ischemic stroke (AIS). This study developed and validated machine learning (ML) models integrating inflammatory biomarkers with clinical indicators to predict post-IVT HT.

methodsIn 1272 IVT-treated AIS patients, the least absolute shrinkage and selection operator (LASSO) regression identified five predictors from 17 variables, which were subsequently utilized to construct eight ML models. The models were trained (70% data) and tested (30% data). Furthermore, external validation conducted on an independent cohort substantiated the generalizability of the optimal model. The SHapley Additive exPlanations (SHAP) method explained feature importance.

resultsLASSO screened five significant predictors: the neutrophil-to-lymphocyte ratio (NLR), admission National Institutes of Health Stroke Scale (NIHSS) score, the Alberta Stroke Program Early CT Score (ASPECTS), blood glucose, and atrial fibrillation. Logistic regression (LR) achieved optimal performance with an AUC of 0.833 internally and 0.842 externally. SHAP analysis prioritized NIHSS as the top contributor, while the nomogram elucidated the variability in HT risk.

conclusionIntegrating NLR with stroke severity and neuroimaging biomarkers enhances the accuracy of HT predictions. The LR-based nomogram provided a practical tool for personalized IVT decisions, emphasizing the prognostic value of systemic inflammation in AIS management.