Early Heart Rate Predicts 3-Month Outcomes in Acute Ischemic Stroke Patients Receiving Intravenous Thrombolysis: A Machine Learning Approach

Mian-Xuan Yao¹Min-Yi Yao¹Jia Gu²Ting Gao^2,3,4Yi-Mian Yuan⁵Yang-Kun Chen^1,4,6YongLin Liu^1,4,7,8*

• ¹Department of Neurology, The Tenth Affiliated Hospital of Southern Medical University (Dongguan People’s Hospital), Dongguan, China
• ²School of Mathematics and Statistics, Huazhong University of Science and Technology, Wuhan, China
• ³Center for Mathematical Science, Huazhong University of Science and Technology, Wuhan, China
• ⁴Guangdong Provincial Key Laboratory of Mathematical and Neural Dynamical Systems,Great Bay University, Dongguan, China
• ⁵Guangdong Medical University, Zhanjiang, China
• ⁶Intelligent Brain Imaging and Brain Function Laboratory (Dongguan Key Laboratory), Dongguan People’s Hospital, Dongguan, China
• ⁷Intelligent Brain Imaging and Brain Function Laboratory (Dongguan Key Laboratory), Dongguan People’s Hospita, Dongguan, China
• ⁸Jinan University, Guangzhou, China

Background The predictive role of early heart rate (HR) dynamics in acute ischemic stroke patients (AIS) receiving intravenous thrombolysis (IVT) remains unclear. This study aimed to evaluate whether HR variability within 24 hours post-IVT predicts early neurological deterioration (END) and 3-month functional outcomes using machine learning. Methods This retrospective analysis included AIS patients without atrial fibrillation (AF) who received IVT at Dongguan People's Hospital between January 2017 and December 2022. Hourly HR metrics (mean HR, SD, coefficient of variation [CV]) were analyzed. Primary outcomes were END (≥ 4-point NIHSS increase within 72 hours) and poor 3-month outcomes (mRS ≥3). Machine learning models were developed and validated via receiver operating characteristic (ROC) analysis. Results Among 381 patients, logistic regression identified NIHSS on admission (OR=1.287, p<0.001), maximum HR (OR=0.956, p=0.023), minimum HR (OR=1.027, p=0.001), and HR SD (OR=1.356, p=0.002) as independent predictors of poor 3-month outcomes. HR CV also showed significance but correlated strongly with SD. A machine learning model integrating onset-to-treatment time, NIHSS, and HR parameters (max/min HR, mean HR, SD) achieved an area under the ROC curve (AUC) of 0.82 for predicting 3-month outcomes. No HR metrics were significantly associated with END. Conclusion In AIS patients without AF, early HR dynamics—particularly maximum HR, minimum HR, SD, and CV—strongly correlate with 3-month functional outcomes after IVT. The machine learning model demonstrated high predictive accuracy, highlighting the potential of real-time HR monitoring for risk stratification and personalized management in thrombolysis-treated AIS patients.