摘要
Abstract
Stroke recurrence risk prediction is a core part of secondary prevention for stroke.The choice of modeling methods directly affects the performance and clinical applicability of the prediction model.This article systematically discusses the common modeling methods of ischemic stroke recurrence risk prediction models at home and abroad and their research progress,including traditional statistical models such as Logistic regression and Cox proportional hazards regression,as well as machine learning models such as Decision tree,Random Forest(RF),Support Vector Machine(SVM),XGBoost and LightGBM,and deep learning models such as BP neural network(BPNN),Convolutional Neural Network(CNN),Long Short-Term Memory Network(LSTM).It focuses on analyzing the advantages,limitations,and applicable scenarios of each model,proposes model optimization strategies,and explores the integration path in clinical decision support system.It is proposed that future research should integrate multimodal data and lightweight design to collaboratively optimize model performance.Through interdisciplinary collaboration to balance the relationship between prediction performance,clinical applicability,and transformation costs,in order to provide ideas for constructing an efficient and practical stroke recurrence prediction model.关键词
卒中/复发/预测模型/建模方法/机器学习/深度学习Key words
stroke/recurrence/prediction model/modeling method/machine learning/deep learning分类
医药卫生
