Research On Non-invasive Continuous Blood Pressure Measurement Method Based On PSO-GRNN Neural Network

Blood pressure(Blood Pressure,BP),as an important physiological parameter of the human body,can judge the heart function and blood vessel health.It can play an important application value in health monitoring,disease warning,curative effect observation,etc.The measurement and monitoring of blood pressure value is The fundamental means and methods for evaluating blood pressure levels,so the research on blood pressure measurement methods and methods and other related technologies is of great significance.The existing blood pressure measurement methods are mainly divided into two categories: invasive blood pressure measurement methods and non-invasive blood pressure measurement methods.The invasive blood pressure measurement method is to insert a special catheter into the subject’s artery and monitor the blood pressure through the connected pressure sensor,thereby realizing continuous blood pressure measurement.Although the measurement data is accurate,this method is not flexible.It is not suitable for daily blood pressure monitoring;non-invasive blood pressure measurement methods are divided into non-invasive intermittent blood pressure measurement methods and non-invasive continuous blood pressure measurement methods according to the measurement continuity.The intermittent measurement method can only measure the time of measurement of the subject The continuous measurement method can measure the blood pressure value of the subject for continuous time.With the development of artificial intelligence(AI),remote monitoring and management of blood pressure based on Internet and AI technology has become a popular research direction.This paper proposes a non-invasive continuous blood pressure measurement method based on the Generalized Regression Neural Network(GRNN)optimized by the PSO(Particle Swarm Optimization Algorithm)algorithm for the problem of non-invasive continuous blood pressure measurement.This method has achieved good results.The main contents of this article include:1.Preprocessing of pulse signal.These include:(1)Remove the noise doped in the pulse signal acquisition process.In this paper,the wavelet transform method is used to remove baseline drift and high-frequency noise.(2)Use the differential threshold method to extract the characteristic points of the pulse wave.Four characteristic points of pulse wave starting point,peak point,descending isthmus point,and dicrotic wave peak point are selected to lay the foundation for follow-up research.(3)Eliminate abnormal pulse data.Due to errors in acquisition or other operations,there is an abnormal pulse cycle signal in the pulse signal.Median filtering is performed on the extracted single pulse cycle,rising branch time,rising branch amplitude,falling branch time,and falling branch amplitude.The upper and lower thresholds are set to eliminate abnormal pulse cycles.When one of the five characteristic indicators is If one or more of them exceed the threshold range,the pulse signal of this period is judged to be an abnormal pulse signal,and it is directly eliminated to prepare for the subsequent establishment of the characteristic matrix.2.Establish a PSO-GRNN neural network non-invasive continuous blood pressure measurement model.The pulse wave data and synchronized blood pressure data in the MIMIC(Multi-parameter Intelligent Monitoring in Intensive Care)database are used as the research object.41 sample data are selected for a total of 99113 pulse data.Establish an11-dimensional pulse wave characteristic parameter matrix,use the particle swarm optimization algorithm to optimize the GRNN neural network,train and determine the model parameters,and finally complete the establishment of the blood pressure prediction model.3.Verify the accuracy of the PSO-GRNN neural network method through comparative experiments.The BP neural network and Elman neural network are selected for experimental comparison and analysis with the method in this paper,and the effect of the PSO-GRNN neural network continuous blood pressure measurement model is verified through experiments due to the other two methods.The pulse signals and synchronized blood pressure data of 5 volunteers were collected for experimental verification,and the final results met the blood pressure measurement effect test standard of the Association for the Advancement of Medical Instrumentation(AAMI).It proves the accuracy and feasibility of continuous blood pressure measurement model based on PSO-GRNN neural network in measuring blood pressure.A non-invasive continuous blood pressure measurement system based on PSO-GRNN neural network is realized.The mpvue framework is equipped with the We Chat applet platform to query and display the blood pressure measurement results on the mobile phone.The Node.js is used in the background to build a Web server to implement algorithm calls and port creation,and to achieve front-end and back-end connections.