渐消记忆递推最小二乘法与曲线拟合法的比较

Comparison Between Two Least Square Algorithm Methods in Respiratory Parameter Measurement

在油酸致急性肺损伤动物模型上,用渐消记忆递推最小二乘(RLS)算法辨识跟踪慢时变呼吸力学参数,并与最小二乘曲线拟合法得出的参数估计结果进行了比较,除肺力学参数外,还对胸—肺系统总的呼吸力学参数也一并进行了估计。结果显示:在注入油酸的四小时期间,RLS与LS算法估计吸、呼时相的胸——肺系统总顺应性(Crs)与肺顺应性(CL)结果基本一致,均显示下降变化,且以Crs为CL的1/2左右。但估计呼吸阻力的结果不一致,LS算法得出的结果不符合生理学先验知识,故R往往可看做是对算法的一种考验。RLS算法精度高、收敛速度快,适于跟踪慢时变的呼吸力学参数。对临床监测,宜选用一室一阶线性呼吸力学模型跟踪呼气相的呼吸力学参数。

The slow time - varying respiratory parameters were traced with recursive least square algo-rithm(RLS) and curve fitting(LS)method on canine acute lung injury models induced by oleic acid. Besides lung mechanics,the total respiratory parameters of chest -lung system were estimated at the same time.The two algorithm methods were also compared. The results showed that during the 4 hour period after oleic acid infusion,the compliance parameters of chest - lung system(Crs)and lung(CL) in inspiratory phase and expira-tory phase estimated by RLS algorithm or LS method were consistent with each other. They all decreased and Crs was about half as large as CL in value. But the results of estimated R parameter did not agree with each other. The results from LS method were not satisfactory and did not accord with physiological knowledge. The RLS algorithm is highly precisive,fast convergent and suitable for tracing slow time - varying respiratory parameters.