用径向基神经网络预测丙戊酸钠血药浓度

Predicting the Plasma Concentration of Sodium Valproate Using Radial Basis Function Neural Networks

应用径向基(RBF)神经网络建立预测丙戊酸钠(VPA)的血药浓度的模型。以患者的性别、年龄、给药间隔、给药剂量、肝功能、肾功能等为输入变量,VPA血药浓度为输出变量,训练RBF网络,获得两者间的关系。当SPREAD值为0.1时,网络模型的预测效果和泛化能力较好。RBF网络用于预测VPA血药浓度的研究是可行的和有效的。

To establish a model for predicting the plasma concentration of sodium valproate （VPA） using radial basis function （RBF） neural networks, RBF neural networks were trained with the data to capture the relationships between the input variables （the patients＇ age, gender, dose, dosing interval, hepatic function, renal function, and etc. ） and the output variable （ VPA plasma concentration）. The model had the better predictive effect and generalization when the SPREAD value was 0.1. It is practical and valid for RBF neural network model to be applied to the study of VPA plasma concentration prediction.