基于人工神经网络的丙戊酸钠个体化给药设计  

Individual administration model for sodium valproate established using artifitial neural network

作　　者：王军[1] 付翠香[1] 吴伟明[1] 王增寿[1] 徐鸿翔[2] 

机构地区：[1]温州医学院附属第二医院药剂科,温州325027 [2]温州医学院药学院,温州325000

出　　处：《中国临床药学杂志》2013年第1期39-42,共4页Chinese Journal of Clinical Pharmacy

摘　　要：目的运用人工神经网络技术构建预测模型，为丙戊酸钠（VPA）临床个体化给药提供参考。方法收集温州医学院附属第二医院200名儿童癫痫患者门诊血药浓度监测数据，分析血药浓度及给药剂量的影响因素，确定建模因子。采用遗传一反向传播算法（GA-BP）建立两层链式预测模型，构建VPA个体化给药的程序并在少量样本中试用。结果初步完成个体化给药的程序编写，30个病例的预测结果与实际血药浓度相比表明，误差值〈15％的有21个，误差为15％-20％的有6个，误差〉20％有3个。结论可以尝试用人工神经网络进行VPA个体化给药的研究。AIM To build a predictive model constructed by artificial neural network, and to provide individual administration reference for sodium valproate. METHODS The cases of the 200 outpatients＇ children with epilepsy were collected. Influence factors on concentration and dosage of sodium valproate were analyzed and model factors were deter- mined. A scheme of two chained dosage prediction models was established by GA-BP algorithm, and some samples were used to validate the model. RESULTS A tentative program for individual administration was compiled. The predicted plasma concentration of sodium valproate for 30 patients demonstrated that the deviations of 21 samples were below 15%, those of 6 samples were between 15% - 20% ,and those of 3 samples exceeded 20%. CONCLUSION A individual ad- ministration research of sodium valproate could be attempted by artificial neural network.

关 键 词：人工神经网络 丙戊酸钠 个体化给药 

分 类 号：R96[医药卫生—药理学]