Pulse Width At Half Amplitude Based Cuffless Continuous Blood Pressure Estimation

Accurate measurement of blood pressure is of great significance to the early prevention,diagnosis and treatment of hypertension and hypertension-related cardiovascular diseases.Currently,there are two main methods for non-invasive continuous blood pressure estimation without a cuff.One is based on machine learning or data-driven methods.This type of method is mainly based on a large amount of data for training.Although blood pressure can be obtained in the end,the specific reasons are difficult to explain,which is clinically unfeasible.The other is a method of physiological analysis based on knowledge about hemodynamics.Its advantage is that every step of the built model is interpretable.This study mainly starts from the second method,focusing on some indicators that are considered to have physiological significance,including pulse width at half amplitude(PWHA),pulse wave systolic area(SA),pulse transit time(PTT)and pulse wave amplitude(AM).On this basis,this paper explores the physiological mechanism model of blood pressure estimation.First,some preliminary explorations were made around PWHA.The linear regression model was used to estimate the blood pressure,and the data of 52 subjects were verified.The results show that the estimation accuracy of systolic blood pressure and diastolic blood pressure is-0.79±6.07 mm Hg and-0.18±3.83 mm Hg,respectively,which is significantly higher than the blood pressure estimation method based solely on PTT,and based on PWHA.The blood pressure estimation accuracy of the fusion of PWHA and PTT is improved compared with the blood pressure estimation accuracy of single PWHA or single PTT.Preliminary results show that PWHA can effectively estimate cuffless continuous blood pressure,which provides a new idea for non-cuff continuous blood pressure estimation.Then,combined with some characteristics of the previous screening,including PWHA,SA,PTT and AM,we adopt a random combination method.Linear regression method for blood pressure estimation.The results show that it is potential to estimate blood pressure when SA is used alone,and at the same time,when the four features are fused together for blood pressure estimation,the effect is the best.For systolic blood pressure and diastolic blood pressure,the estimation accuracy is 0.29±4.97 mm Hg and-0.14±3.06 mm Hg respectively,and the mean absolute difference is 3.83 mm Hg and 2.32 mm Hg respectively.Finally,combined with Poiseuille’s formula and pulse wave contour analysis,three physiological mechanism models were sequentially deduced,and the data of 17 healthy subjects were used for model validation.The results show that model three works the best.The estimation accuracy of systolic blood pressure,diastolic blood pressure and mean blood pressure estimated by this method were 1.85±15.35 mm Hg,1.41±15.52 mm Hg and1.56±15.40 mm Hg,respectively,and the mean absolute differences were 11.82 mm Hg,11.85 mm Hg and 11.78 mm Hg,respectively.In summary,the work in this paper lays a good foundation for exploring mechanistic models of blood pressure estimation.