用人工神经网络建立缺血性脑卒中复发的预测模型

Building Ischemic Stroke Recurrence Prediction Model by Using Artificial Neural Networks

目的利用BP人工神经网络和多因素logistic回归建立缺血性脑卒中患者复发的预测模型,为临床医生提供一种简单、高效、准确的评估缺血性脑卒中患者是否复发的方法。方法应用logistic回归模型对资料进行单因素筛选,将筛选出有统计学意义的指标进行BP神经网络和多因素logistic回归分析,建立缺血性脑卒中患者复发的预测模型,并对两个模型进行比较。结果应用BP神经网络和logistic回归模型建模,对测试集样本预测,BP神经网络和logistic回归模型预测正确率分别为84.6%和81.7%,ROC曲线下面积分别为0.787和0.729,说明BP神经网络模型预测性能优于logistic回归模型。结论人工神经网络模型预测效果优于logistic回归模型。

Objective To establish ischemic stroke recurrence prediction model based on BP artificial neural network, provides clinicians with a simple, efficient, accurate assessment of patients with ischemic stroke recurrence. Methods Using logistic regression model to univariate analy- sis, and filter out the significant indicators for the BP neural network and lo- gistic regression multivariate analysis, to establish the ischemic stroke recur- rence prediction model, and compared two models. Results Establish the BP neural network and logistic regression model, BP neural network and lo- gistic regression＇s prediction accuracy were 82.6% and 75.1%, The area under the ROC curve were 0. 875 and 0. 880, BP neural network model has a better prediction accuracy than the logistic regression model. Condusion Artificial neural network model is better than the logistic regression mod- el in prediction effect.