人工神经网络技术预测癫痫患儿服用丙戊酸后体内药物浓度  

Prediction the plasma concentration of valproate in patients with epilepsy by artificial neural network

作　　者：邱枫[1] 何晓静[1] 肇丽梅[1] 孙亚欣[1] 

机构地区：[1]中国医科大学附属盛京医院药学部,沈阳110004

出　　处：《中国临床药理学杂志》2012年第2期96-98,共3页The Chinese Journal of Clinical Pharmacology

基　　金：辽宁省科学技术计划基金资助项目(2009225020)

摘　　要：目的利用人工神经网络技术预测癫痫患儿服用丙戊酸后体内药物浓度。方法收集200例癫痫患儿服用丙戊酸后血药浓度监测结果、身高、体重及监测当日肝肾功能等15项相关指标,根据神经网络和遗传优化反向传播算法的基本原理,构建丙戊酸血药浓度预测模型,并用该浓度预测模型进行样本预测分析。结果 50个病例样本的预测结果表明,与实际测定浓度相比,误差小于10%的有29个浓度,误差在10%～15%的有10个浓度,误差在15%～20%的有7个浓度,误差大于20%的有4个浓度。误差小于15%的比率是78%。人工神经网络预测的血药浓度和实际测定浓度之间的相关系数为0.9476。结论用人工神经网络技术预测癫痫患儿服用丙戊酸后的血药浓度是可行的;有待将其广泛应用于个体化给药设计。Objective To predict plasma concentration of valproate in patients with epilepsy by artificial neural network（ANN） simulator.Methods A data set of 15 physiological measurements for 200 patients was used to develop the model.Predictive model on plasma concentration of valproate was based on neural network and GA-BP.Then,samples were forecasted using the predictive model.Results Plasma concentration of valproate from 50 patients demonstrated that the deviations of 29 points were less than 10 %,that of 10 points were between 10% and 15 %,that of 7 points were between 15 % and 20 %,and that of 4 points exceed to 20%,compared with determination concentration.The ratio of less than 15% error is 78 %.The correlation coefficients between determinated and predicted values obtained by ANN prediction using standardized data sets were 0.9476.Conclusion Prediction of plasma concentration of valproate in patients by ANN is feasible,which will be widely used in individualized dosage design.

关 键 词：人工神经网络 丙戊酸 血药浓度 预测 

分 类 号：R971.6[医药卫生—药品] R969.3[医药卫生—药学]