摘要
目的运用人工神经网络技术构建预测模型,实现环孢素A(CsA)的个体化给药设计。方法收集肾移植患者临床数据,分析CsA血药浓度及给药剂量的影响因素,确定建模因子。采用遗传-反向传播算法(GA-BP)建立两层链式预测模型,构建CsA个体化给药的程序并少量样本试用。结果初步完成个体化给药的程序编写,10个病例(总共30次)的预测表明,绝大多数次(23/30)的预测给药剂量与实际给药剂量的误差小于15%,超过半数次(17/30)的误差小于10%。结论可以尝试用人工神经网络进行CsA个体化给药的研究。
OBJECTIVE To build a predictive model constructed by artificial neural network,and to adjust administration of cyclosporine A（CsA） individually.METHODS The cases of kidney transplantation were collected.Influence factors on concentration and dosage of cyclosporine A were analyzed and model factors were determined.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 predictive doses of cyclosporine A for 10 patients（30 points in all） demonstrated that the deviation of majority of them（23/30）was below 15% and some of them（17/30） was below 10%,compared with actual doses.CONCLUSION An individual administration of cyclosporine A could be attempted by artificial neural network.
出处
《中国药学杂志》
CAS
CSCD
北大核心
2010年第12期927-930,共4页
Chinese Pharmaceutical Journal
关键词
人工神经网络
环孢素A
个体化给药
artificial neural network
cyclosporine A
individual administration
