Noninvasive Measurement Of Dynamic Blood Pressure Waveform Based On Pulsatile Flow Simulation Device

Blood pressure waveform signals contain rich information of cardiovascular diseases,the continuous monitoring of blood pressure waveform is of great significance to the evaluation of cardiovascular status.At present,noninvasive blood pressure measurement is an important means of cardiovascular health monitoring.However,at present,not all noninvasive blood pressure measurement methods can detect continuous blood pressure waveform,and the relationship between threedimensional deformation of blood vessel wall caused by blood pressure and ambulatory blood pressure waveform has not been deeply studied.The threedimensional deformation of blood vessel wall is the direct result of the change of blood pressure,which is closely related to blood pulsation.Based on the self-developed pulsatile flow simulation device,this thesis uses binocular vision method to study the internal relationship between the threedimensional deformation of vascular wall and the waveform of intravascular pressure,and reveals the mechanism of the spatial characteristics of vascular wall caused by the change of intravascular pressure.Firstly,based on the formation mechanism of pulse and blood pressure and the principle of hemodynamics,a pulsatile flow simulation device which can produce multiple types of threedimensional deformation of vascular wall and corresponding pressure waveform in the tube is developed.Through servo motor,hydraulic device,silicone tube and other equipment and materials,build platform that can simulate blood circulation.By simulating human heart activity,the platform driver is designed;Select appropriate devices to simulate hemodynamic parameters,so that the device can produce multiple types of spatiotemporal pulse data and blood pressure waves.The device can simulate the pulsation of blood vessels,which is conducive to study the internal relationship between threedimensional pulse displacement and blood pressure waveform,and provide a reference platform for noninvasive measurement of blood pressure waveform.Secondly,the threedimensional deformation of blood vessel wall generated by binocular vision acquisition platform is regarded as pulse image,which contains rich spatiotemporal pulse data;Use the pressure transmitter to synchronously collect the blood pressure waveform in the tube.Based on the principle of binocular vision ranging,the multidimensional pulse wave of spatiotemporal pulse data is extracted from the pulse image.Ensemble Empirical Mode Decomposition(EEMD),Complete Ensemble Empirical Mode Decomposition(CEEMDAN)and Variational Mode Decomposition(VMD)are used to denoise the simulated pulse wave and blood pressure wave.The algorithm is set so that VMD can automatically obtain the number of optimal intrinsic mode function,and the denoising effects of the three methods are evaluated by spectrum analysis.The experimental results show that VMD method can reduce the medium and high frequency noise energy of pulse wave to less than 0.005,and can retain more low frequency pulse and blood pressure signals,and the denoising effect is the best.Thirdly,the blood pressure waveform measurement model is constructed through Linear Regression(LR),Support Vector Regression(SVR)and Random Forest(RF)Regression to establish the internal relationship between spatiotemporal pulse data caused by threedimensional dynamic displacement of blood vessel wall and blood pressure wave.The measurement effects of three blood pressure waveform measurement models were analyzed by using Correlation Coefficient(R)and Mean Square Error(MSE)as evaluation indexes.The experimental results show that the blood pressure waveform measurement model constructed by RF has higher accuracy.Finally,aiming at the problem of information loss in the learning process of regression method and the differences in temporal and spatial characteristics of different types of pulse waves,different window lengths are set for different types of spatiotemporal pulse data,and the blood pressure waveform measurement model under multi window length is constructed.Since the training data are pulse and blood pressure time series signals,the improved network of recurrent neural network(RNN)is adopted,namely Long-Short Term Memory(LSTM),Bi-directional Long-Short Term Memory(Bi LSTM)and Gated Recurrent Unit(GRU)network,construct a blood pressure waveform measurement model with long-term memory function.The experimental results show that Bi LSTM can learn multitype and multidimensional pulse waves in two directions,and the average R between the measurement results and the true value is 0.968,which is higher than the measurement results of LSTM and GRU.