论文部分内容阅读
目的探讨使用BP神经网络技术用于评估肾小球滤过率的意义。方法根据神经网络原理,进行BP网络建模,并进行训练,直到均方误差(MSE)<10-8。选择我科慢性肾病患者1 330例,分别使用MDRD简化方程和训练好的BP网络对肾小球滤过率进行计算,结果使用方差分析和线性回归分析,比较两种方法的优缺点。结果两种方法方差分析P=0.590,均值比较差异无统计学意义。线性相关系数0.999(P<0.001),两种方法计算结果高度相关。结论BP神经网络可以用于评估肾小球滤过率,为探索更优的评估方程提供了一种不同于回归方程的方法。
Objective To explore the significance of using BP neural network to evaluate glomerular filtration rate. Methods According to the principle of neural network, BP network modeling and training until the mean square error (MSE) <10-8. We selected 1 330 patients with chronic kidney disease in our department, calculated the glomerular filtration rate using MDRD simplified equation and trained BP network respectively. The results were compared using ANOVA and linear regression analysis, and the advantages and disadvantages of the two methods were compared. Results The two methods ANOVA P = 0.590, the mean difference was not statistically significant. The linear correlation coefficient was 0.999 (P <0.001), and the results of the two methods were highly correlated. Conclusion BP neural network can be used to evaluate the glomerular filtration rate, to explore a better evaluation of the equation provides a method different from the regression equation.