论文部分内容阅读
目的:探讨人工神经网络预测早期先兆流产结局的价值。方法:将136例符合条件的孕早期先兆流产孕妇随机分为观察组(50例)和对照组(86例)。观察组分别选用生化参数、超声参数和联合参数作为输入参数来构建3种不同的神经网络,以86例对照组来分别测试3种网络的稳定性和误差。结果:联合参数法准确率最高为85.7%,生化参数法为81.9%,超声参数法为78.7%。结论:使用人工神经网络对早期先兆流产结局的预测具有很好的前景。选择合适的参数建立网络可更加提高预测的准确性。
Objective: To explore the value of artificial neural network in predicting outcome of early threatened abortion. Methods: 136 pregnant women with threatened abortion in early pregnancy were randomly divided into observation group (50 cases) and control group (86 cases). In the observation group, three kinds of neural networks were selected using biochemical parameters, ultrasonic parameters and joint parameters as input parameters respectively. The stability and error of the three kinds of networks were tested by 86 control groups respectively. Results: The accuracy of joint parameter method was up to 85.7%, biochemical parameter method was 81.9% and ultrasonic parameter method was 78.7%. Conclusion: Artificial neural network prediction of early threatened abortion outcome has good prospects. Choosing the right parameters to build a network can increase the accuracy of predictions.