基于脉搏波到达时间的3种血压计算模型的对比研究

A comparative study of three BP estimation models based on PAT

目的连续血压监测对心血管疾病的早期诊断和预防具有重要意义。对基于脉搏波到达时间(PAT)的不同血压计算模型在同一校准方法下进行对比研究,寻找更加精确的血压模型。方法选择11例受试者,其中男性8例,女性3例;年龄21.5~29.3岁,平均年龄25.3岁;身高160.6~176.3 cm,平均身高168.5 cm。要求在5 min内以一定的速度从1楼徒步到12层,使血压明显升高。采集受试者的光电容积脉搏波信号(PPG)和心电信号(ECG),同时测量的袖带压作为参考血压值。采用PAT#1(基于质量守恒方程和动量守恒方程的弹性管道模型)、PAT#2(从Moens-Korteweg方程和指数动脉弹性模型中分离所得)、PAT#3(从Moens-Korteweg方程和Bramwell-Hill方程中所提取)模型,计算3种模型的血压值。结果 3种模型中收缩压(SBP)的相关系数高于舒张压(DBP)的相关系数。其中:PAT#1模型的SBP和DBP相关系数最小(rSBP=0.833,P=0.002 4;rDBP=0.735,P=0.003 5);另外两种模型之间差异无统计学意义(P> 0.05)。PAT#1对应的3个误差统计量均最大,说明其精确度最低。对于SBP和DBP,PAT#1对应的ME与PAT#2、PAT#3相比,差异都具有显著统计学意义(SBPPAT#1 vs PAT#2:F=16.54,P <0.001;DBPPAT#1 vs PAT#2:F=12.76,P <0.001;SBPPAT#1 vs PAT#3:F=17.37,P <0.001;DBPPAT#1 vs PAT#3:F=9.31,P <0.001)。结论基于血管弹性模型的血压计算模型在连续血压监测上更加准确,误差更小。

Objective To compare and study different blood pressure(BP) calculation models based on pulse arrival time(PAT)under the same calibration method, and find a more accurate BP evaluation model, because of continuous BP monitoring is of great significance to the early diagnosis and prevention of cardiovascular diseases. Methods Eleven subjects were selected,which included 8 males and 3 females, aged 21.5-29.3 years old with mean age of 25.3 years old, height 160.6-176.3 cm with mean height of 168.5 cm. All of the subjects were required to walk from the first floor to the 12 th floor at a certain speed within 5 minutes to increase BP significantly. The photoplethysmographic(PPG) and electrocardiogram(ECG) of the subjects were collected, and the measured cuff pressure was used as the reference BP value. The PAT#1 model(elastic pipe model based on conservation of mass and momentum equations), PAT#2 model(from Moens-Korteweg equation and exponential arte-rial elastic model) and PAT#3 model(from Moens-Korteweg equation and Bramwell-Hill equation model) were chosen, and BP values of the three models were calculated. Results The correlation coefficient of systolic blood pressure(SBP) was higher than that of diastolic blood pressure(DBP) in 3 models. The correlation coefficient between SBP and DBP of PAT#1 model was the smallest(rSBP= 0.833, P = 0.002 4;rDBP= 0.735, P = 0.003 5);the difference between the other two models was no statistically significant(P > 0.05). All of the 3 error statistics corresponding to PAT#1 were the largest, which showed the lowest ac-curacy. For SBP and DBP, the mean error(ME) corresponding to PAT#1 was compared with that of PAT#2 and PAT#3, respec-tively, and the differences were statistically significant(SBPPAT#1 vs PAT#2: F = 16.54, P < 0.001;DBPPAT#1 vs PAT#2: F = 12.76, P < 0.001;SBPPAT#1 vs PAT#2: F = 17.37, P < 0.001;DBPPAT#1 vs PAT#2: F = 9.31, P < 0.001). Conclusion It is demonstrated that the BP estimation model based on vascular elasticity model is more accurate with smaller errors in continuous BP monitoring.