Research On Cuffless Blood Pressure Measurement Method For High BMI Population

Blood pressure values are closely related to the function of the human cardiovascular system and the elasticity of the arteries,so they can be used to show whether there is a pathology in its function.In recent years,as people’s standard of living improves,irregular lifestyle habits have led to an increasing number of people with high Body Mass Index(BMI)and abnormal blood pressure,especially hypertension,which poses a risk of cardiovascular diseases such as coronary heart disease or myocardial infarction.Studies have shown that continuous ambulatory blood pressure measurement allows people to be informed of their blood pressure in real time,which is significant for the prevention,detection and treatment of hypertension in a timely manner.Invasive blood pressure measurement is an important method for continuous ambulatory blood pressure measurement and is known as the "gold standard" blood pressure value,but it can cause trauma to the patient’s internal tissues and is generally used only in intensive care scenarios.Therefore,blood pressure measurement based on non-invasive means is the focus of most research,and noninvasive blood pressure measurement devices are generally classified into two types:cuffed and non-cuffed.The current more traditional is blood pressure measurement through a cuff device,but it can cause discomfort during pressurization and is not suitable for continuous dynamic monitoring and other anytime,anywhere measurements.As an important physiological waveform of the human body,the waveform pattern contains a lot of information related to the human body mechanism and physiology.By measuring the pulse waveform and designing an algorithm to analyze the waveform,the blood pressure measurement can be realized,and thus the cuff device can be removed.Research on blood pressure measurement methods based on pulse waves has become a focus in recent years.At present,many pulse wave measurement devices require the installation of cardiac leads or finger clip devices,and the operation process is cumbersome,so researchers generally do not collect their own data for blood pressure measurement studies,but use relevant data from the publicly available Medical Information Mart for Intensive Care(MIMIC).However,the distortion of pulse wave signal in this dataset is serious,feature extraction is complicated,and it lacks physiological information such as height and weight,which can only perform general type of blood pressure modeling and increases the risk of inaccurate blood pressure measurement in people with high BMI.In this paper,we design a hardware module that can realize pulse wave acquisition by reflective photoelectric sensors,and use the module to acquire pulse wave data and establish a new pulse wave feature parameter-blood pressure dataset.Then a Stacking integrated blood pressure measurement model for people with high BMI is designed and validated using body mass index as the basis for population classification.The main research of this paper is as follows.Firstly,by selecting the photoelectric sensor,analog front-end,microcontroller,power supply and Bluetooth module,and then designing the hardware circuit for each module in turn.According to the circuit schematic,PCB technology is used to connect all modules together,and finally a pulse wave acquisition module with small size,light weight and low power consumption is designed.Secondly,volunteers of all ages and different heights and weights in good cardiovascular health were recruited to design an experiment for pulse wave data acquisition.Then the wavelet transform and Butterworth low-pass filter are used to filter out the high-frequency noise,baseline drift and burr interference contained in the original signal to obtain a distinctive pulse wave signal.The adaptive detection algorithm combining threshold and difference,second-order difference method,and wavelet transform method are used to detect all the feature points,and then a series of feature parameters such as amplitude parameter,ratio parameter,slope parameter,time parameter,and area parameter are extracted,and combined with vital sign information such as height,age,and weight to establish a new pulse wave data set.Again,The dataset was divided into non-high BMI population(BMI < 25)and high BMI population(BMI > 25)sub-datasets according to BMI values,and a single linear blood pressure measurement model and a single non-linear blood pressure measurement model were developed for different populations using BMI as a new characteristic parameter,and then each model was evaluated,and it was found that the single model was generally more accurate in measuring blood pressure for the high BMI population than for the non-high population.Finally,given that the overall accuracy of a single regression model for different populations is not high,the optimal single nonlinear regression model(K-nearest neighbor,extreme random forest,light GBM regression model)is chosen as the first layer learner,and the linear regression model is used as the second layer learner to build the Stacking integrated regression-based blood pressure measurement model.Using the Stacking model to train and test different populations,and then using the blood pressure measured by a mercury sphygmomanometer as the "standard value" for cuff free blood pressure,the measured values were comprehensively evaluated.The experimental results are as follows: In the population with non high BMI,the measurement accuracy of the Stacking model for systolic and diastolic blood pressure(RMSE/MEA)is 6.6204 mm Hg/5.0010 mm Hg and 4.6161 mm Hg/3.5115 mm Hg,respectively;In people with high BMI,the measurement accuracy of systolic and diastolic blood pressure,RMSE/MEA,is 6.300 mm Hg/4.8976 mm Hg and 4.3657 mm Hg/3.2334 mm Hg,respectively.The accuracy of systolic and diastolic blood pressure measurements meet the Association for the Advancement of Medical Instrumentation(AAMI)standards.The pulse wave acquisition module designed in this article has a simple measurement method.In daily life,it is only necessary to place a finger on a sensor to complete the acquisition of the photoconductive pulse wave signal,completely breaking away from the constraints of finger clip equipment and electrocardiographic leads.The built Stacking integrated regression blood pressure measurement model effectively solves the problem of low accuracy of existing algorithms for blood pressure measurement in people with high BMI,It provides greater possibility for the majority of obese patients to conduct daily blood pressure monitoring.