基于小波神经网络时间序列模型预测血药浓度的研究

基于小波神经网络时间序列模型预测阿司匹林血药浓度，来评价模型的适应性。方法 对4组家兔进行灌胃，分别于0.15 h、0.25 h、0.5 h、1.0 h、1.5 h、2.0 h、2.5 h、3.0 h、3.5 h、4.0 h、6.0 h、13.0 h、22.0 h时间点获取血药浓度数据，利用计算机软件MATLAB对其中3组实验数据进行网络训练，利用训练好的网络对剩下的1组血药浓度数据进行预测。之后联合药代动力学，判断其房室模型和药代动力学特点。结果 模拟仿真结果与实际数据基本相符合，网络的绝对平均误差范围在0.3%~5.39%，在统计学允许误差范围之内。联合药代动力学仿真证明了阿司匹林的血管外给药药动学特点是二室模型。结论 小波神经网络时间序列模型在预测阿司匹林血药浓度时具有较好的拟合能力和优良的预测能力，同时与药代动力学的结合更为现代临床药理的研究起到积极的推动作用。

Prediction of Plasma Concentration Based on Wavelet Neural Network Time Series Model

Abstract Objective To evaluate the adaptability of wavelet neural network time series model through predicting the concentration of aspirin in blood. Methods 4 groups of rabbits were fed with aspirin, and plasma concentration data were obtained in 0.15 h, 0.25 h, 0.5 h, 1.0 h, 1.5 h, 2.0 h, 2.5 h, 3.0 h, 3.5 h, 4.0 h, 6.0 h, 13.0 h, 22.0 h time points. Then through the software MATLAB, 3 groups of experimental data were trained by network, and the trained network was used to predict the plasma concentration data of the 1 group. Its characteristics of the compartment model and pharmacokinetic were determined by combination of pharmacokinetics. Results The simulation results were consistent with the actual data, and the absolute mean error of the network was in the range of 0.3%~5.39%. The pharmacokinetics of aspirin in two compartment model was proved by combination of pharmacokinetic simulation. Conclusion Wavelet neural network time series model in predicting the plasma concentration of aspirin are with good fitting capability and excellent predictive ability, at the same time with the combination of pharmacokinetics more plays a positive role in promoting modern clinical pharmacology research.