论文部分内容阅读
目的应用反向传播神经网络建立癫痫儿童的血清丙戊酸浓度预测模型。方法收集122例癫痫儿童的临床相关资料,根据资料获取的难易程度,应用反向传播神经网络建立建模因子分别为6及12的模型Ⅰ和模型Ⅱ,比较两种模型的预测效能。并利用模型Ⅱ前瞻性预测12例癫痫儿童的血清丙戊酸浓度。结果模型Ⅰ和模型Ⅱ测试组的预测值与实测值的相关系数分别为0.945、0.986;均方预测误差平方根(RMSE)分别为5.864、2.998;平均预测误差百分率分别为0.42%、-0.59%(P>0.05);平均绝对误差百分率分别为7.67%、4.40%(P<0.05)。前瞻性预测(12例)其误差百分率在±5%以内的7例,±5%～±10%的3例,±10%～±20%的1例,大于20%的1例。结论模型Ⅱ对癫痫儿童血清丙戊酸浓度的预测精准度优于模型Ⅰ。所建立的针对儿童的神经网络模型能够有效预测癫痫儿童的血清丙戊酸浓度,可指导临床个体化给药。
Objective To establish a predictive model of serum valproic acid concentration in children with epilepsy using backpropagation neural network. Methods The clinical data of 122 children with epilepsy were collected. Based on the ease of data acquisition, models Ⅰ and Ⅱ with modeling factors of 6 and 12 were established by using backpropagation neural network. The predictive power of the two models was compared. And the use of model Ⅱ prospective prediction of 12 cases of epilepsy serum valproic acid concentrations. Results The correlation coefficients between predicted and measured values of model Ⅰ and model Ⅱ test groups were 0.945 and 0.986, respectively. The RMSE of mean square prediction error was 5.864 and 2.998, respectively. The average prediction error was 0.42% and 0.59% respectively P> 0.05). The mean percentage of absolute error was 7.67% and 4.40% respectively (P <0.05). In the prospective prediction (12 cases), 7 cases were within ± 5% error rate, 3 cases were ± 5% ± 10%, 1 case was ± 10% ± 20% and 1 case was more than 20%. Conclusion The prediction precision of serum valproic acid concentration in model Ⅱ is better than that of model Ⅰ in children with epilepsy. The established neural network model for children can effectively predict serum valproic acid concentration in children with epilepsy and can guide clinical individualized administration.