摘要
目的:建立人工神经网络用于估算他克莫司血药浓度。方法:收集26例肝移植患者口服他克莫司的94份全血浓度数据,采用遗传算法配合动量法优化网络参数,建立人工神经网络。结果:人工神经网络平均预测误差(MPE)与平均绝对预测误差(MAE)分别为(-0.11±2.81)ng/mL和(2.14±1.72)ng/mL,78.6%血药浓度数据绝对预测误差≤3.0ng/mL。多元线性回归MPE与MAE分别为(0.56±2.70)ng/mL和(2.15±1.63)ng/mL,9例次(9/14,64.3%)绝对预测误差≤3.0ng/mL。人工神经网络准确性及精密度优于多元线性回归。结论:人工神经网络预测可用于预测他克莫司血药浓度,指导个体化给药。
AIM: To establish a prediction method for tacrolimus concentration in liver transplantation recipients by artificial intelli- gence. METHODS: 94 tacrolimus concentration samples from 26 Chinese liver transplantation recipients were collected. Artificial neural net- work (ANN) was established after network pa- rameters were optimized by using momentum method combined with genetic algorithm. Fur- thermore, the performance of ANN was com- pared with that of multiple linear regression (MLR). RESULTS: With ANN method, the lev- els of mean prediction error(MPE) and mean ab- solute prediction error (MAE) were ( -- 0.11 ± 2.81) ng/mL and(2.14±1.72) ng/mL, respec- tively. The absolute prediction error of 78.6% of testing data sets was less than 3.0 ng/mL. The levels of MPE and MAE were(0.56±2.70) ng/mL and(2.15 ± 1.63) ng/mL respectively, with MLR method. The absolute prediction er- ror of 64.3% of testing data sets was less than 3.0 ng/mL. Accuracy and precision of ANN was superior to that of MLR. CONCLUSION: ANN is suitable to predict tacrolimus concentration.
出处
《中国临床药理学与治疗学》
CAS
CSCD
2012年第7期791-796,共6页
Chinese Journal of Clinical Pharmacology and Therapeutics
关键词
他克莫司
肝移植
人工神经网络
Tacrolimus
Liver transplanta- tion
Artificial neural network