Non-invasive Continuous Blood Pressure Detection Technology Based On Multi-wavelength Photoplethysmography

Blood pressure is one of the important physiological indicators of the human body,and hypertension is the main cause of cardiovascular disease.Continuous blood pressure monitoring is more predictive of cardiovascular mortality.The problems of continuous blood pressure detection in recent years lie in the inaccurate calculation method of pulse wave transit time(PTT)and the incomplete consideration of blood pressure related factors,especially the systemic vascular resistance(SVR),which does not have continuous measuring method so far,and thus hardly taken into account by researchers in blood pressure calculation models.Therefore,tracking SVR through measurable signals on the body surface is of great significance for continuous blood pressure monitoring.Therefore,this paper introduces multi-wavelength photoplethysmography(PPG),and proposes a multi-wavelength PPG algorithm based on the least mean square(LMS)algorithm to extract the arteriolar pulse transmit time(a PTT)to track SVR.However,this method lacks experimental evidence that the a PTT can reflect SVR.Therefore,this paper also proposes an experimental verification method for the relationship between them.Numerous studies have shown a strong link between the sympathetic nervous system and hypertension.Sympathetic nerve activity affects blood pressure by constricting arterioles,and then increases SVR.In this paper,cold stimulation experiments and emotional stimulation experiments were designed to verify the relationship between a PTT and SVR.In recent years,machine learning methods have become a research hotspot in various fields,and many researchers have also tried to apply them to the problem of blood pressure detection.The machine learning model is built from a large amount of data,with strong generalization ability and high accuracy.However,machine learning-based continuous blood pressure detection methods usually lack physiological interpretability and cannot make targeted improvements to blood pressure models.Therefore,we will build a continuous blood pressure detection model based on machine learning methods,and then verify the importance of a PTT.The main work and results of this paper are as follows:1.A multi-wavelength photoplethysmography algorithm based on LMS is proposed to obtain the a PTT.In this paper,the measurement equipment developed by the laboratory can simultaneously collect multi-wavelength PPG signals,electrocardiograph(ECG)and galvanic skin response(GSR).Since the signal obtained by the traditional PPG method is actually the superposition of the reflection signals of multiple blood vessel layers to light,if it is used as a reflection of the elasticity of the aorta and aorta wall,there will be influences from other blood vessel layers,especially arterioles.In this paper,the arterial and capillary PPG signals are reconstructed in the multi-wavelength PPG signal,and the influence of other blood vessels is excluded,so as to accurately extract the a PTT signal and the artery PTT signal.2.The relationship between a PTT and SVR was demonstrated by cold stimulation experiments and emotional stimulation experiments.In the cold stimulation experiment,the laser Doppler method was used to monitor blood perfusion while the self-developed measuring equipment simultaneously collected multi-wavelength PPG.In the emotional stimulation experiment,the self-developed measurement equipment simultaneously collected multi-wavelength PPG,ECG and GSR.Subsequently,the a PTT was calculated using our proposed LMS-based multi-wavelength algorithm in this paper.In the cold stimulation experiment,in the recovery period after cold stimulation,a PTT decreased with an average slope of-0.2080,while blood perfusion increased with an average slope of 0.7046.The results of the cold stimulation experiments illustrate the relationship between the decrease in a PTT and the decrease in SVR.In emotional stimulation experiments,20 healthy subjects watched movie clips to stimulate sympathetic nerves,and then dynamic time warping(DTW)distance was used to evaluate the correlation between GSR and a PTT.During emotional stimulation,the median dynamic time-warped distance between a PTT and GSR was significantly smaller than the DTW distance between arterial PTT and GSR in70% of participants.The results of emotional stimulation experiments illustrate that a PTT is associated with sympathetically driven SVR.Two experimental results demonstrate that the continuously measurable parameter a PTT can well track SVR.3.Build a continuous blood pressure detection model and verify the improvement of a PTT on the blood pressure model.We used the collected data to construct a continuous blood pressure detection model,among which 74 were hypertensive and 263 were normotensive.After data preprocessing,feature selection,model fitting,and parameter optimization,we made the model meet the AAMI standard.With the participation of a PTT,mean absolute error in systolic blood pressure calculation was reduced by 0.3435 mm Hg in hypertensive people.After the feature importance analysis of the model,it was found that the a PTT is an important feature in the waveform features.At the same time,the importance of a PTT in the prediction of systolic blood pressure in hypertensive population is also greater.We can conclude that a PTT can improve the detection of systolic blood pressure in hypertensive population.