Measurements Of Pulse Information And Blood Pressure Waveform Based On Binocular Vision

According to the statistics in summary of Chinese Cardiovascular Disease Report2016,cardiovascular disease mortality accounted for the first of resident disease death,and showed an upward trend year by year.So the prevention and treatment of cardiovascular disease have become the urgent problems.Pulse beats and change of blood pressure are generated by the heart beats.Clinical studies have shown that some abnormal information has been implied in pulse signal and arterial blood pressure waveform in the early stage of cardiovascular disease.If the pulse information can be fully and accurately detected,and the blood pressure waveform can be noninvasively and continuously monitored,more physiological and pathological information will be obtained,which will contribute to the effective prevention,diagnosis and treatment of cardiovascular disease.And it is of great significance to clinical medicine and basic medicine.Based on the review of research status at home and abroad about the detection and analysis of blood pressure and pulse signal,the advantages and disadvantages of the existing methods were analyzed.And it was concluded that the pulse information detected by the existing methods was not comprehensive enough,and the noninvasive measurement methods of blood pressure waveform still had many defects.Thus,the research idea of detecting pulse signal based on binocular vision measurement principle,and then realizing noninvasive measurement of blood pressure waveform was put forward in this dissertation.First,a binocular vision pulse detection system was developed.In the system,pulse of the radial artery is transformed into the deformation of probe thin film by a flexible probe,and the changes in thin film are captured with two industrial cameras,which makes the pulse beats visible.In this way,the full detection of pulse signal based on image came true.Aiming at the shortcomings of the image acquisition software that comes with the used camera,based on the Application Programming Interface of the camera,the Dual-camera Synchronous Acquisition Software was developed in the Microsoft Visual C++ 2008 environment.With the assistance of a dual-port Gigabit Ethernet card and a MP425 data acquisition module,using the camera’s external trigger mode,stable synchronous acquisition of two cameras and automatic continuous image saving were achieved.A suitable high precisioncalibration plate was selected and the self-made backlight was used to assist the Zhang’s calibration method to calibrate the binocular vision system with small field of view and small working distance.The calibration result was of high precision and met the requirement of pulse amplitude measurement.Second,multi-point pulse waves were extracted based on the dynamic pulse image sequences acquired by the binocular vision pulse detection system.According to the characteristic that the deformation of probe thin film caused by pulse beats leads to the change of image gray value between frames,the pulse wave was extracted using several basic image statistics.After comparing the results,the multi-point pulse wave extraction method based on image entropy was proposed.This method has high efficiency and good waveform stability.It can accurately obtain the pulse beats trend,and the waveform is better than that of the method based on the area change of feature pattern.In order to obtain the actual amplitude of pulse beats,another multi-point pulse wave extraction method was proposed based on longitudinal displacements of feature points.The centroids of feature patterns in pulse images were extracted as the feature points,then the concept of Gong-shape scan detection is introduced to match the feature points in the left and right images.The spatial positions of feature points were calculated based on binocular vision theory,and then the longitudinal displacements were extracted.According to the changes of longitudinal displacements frame by frame,the multi-point pulse waves which can accurately reflect the actual pulse amplitude and pulse beats trend were obtained,and the waveform effect was better than those of the methods based on the area change of feature pattern and image entropy.In addition,the accuracy and effectiveness of this method were tested and verified by static and dynamic verification methods.On the basis of the longitudinal displacements of multiple feature points,the dynamic three-dimensional pulse shape was obtained.According to the spatial positions of multiple feature points at the same time,the three-dimensional shape of probe thin film was reconstructed and the result was consistent with the reality.After the deformation of probe thin film under the force of pulse beats was analyzed,the thin film deformation caused by pulse beats at a moment was reconstructed based on the pulse amplitudes of multiple feature points at the same time.Then,based on pulse image sequence,the dynamic three-dimensional pulse shape changing with time was obtained,which vividly presented the pulse beating of the radial artery.Furthermore,the pulse width and pulse length were calculated based on the dynamic three-dimensional pulse shape using the calculation method defined according to theirbasic concepts.Finally,the blood pressure waveform was obtained based on pulse wave.A blood pressure simulation platform to simulate human blood circulation system was set up for the comparison and analysis of the relationship between the radial displacement waveform of vessel wall and the change of pressure in vessel.And the conclusion was that the signal waveforms detected by binocular vision pulse detection system and pressure transmitter were of good consistency.Accordingly,the research method of blood pressure waveform extraction based on pulse wave was put forward.After analyzing the linear relationship between pressure in the vessel and radial displacement of the vessel wall,the mathematical expression for calculating blood pressure from pulse amplitude was derived.The pulse center was determined as the best position to obtain blood pressure waveform,then the blood pressure waveform was preliminarily estimated.The simple linear correlation analysis method was used to determine the pulse parameters that have strong correlation with blood pressure.Systolic and diastolic blood pressure were obtained by multiple linear regression analysis method,and then used to further correct the blood pressure waveform.In addition,with the help of blood pressure simulation platform,the prediction method of pressure waveform based on BP neural network was preliminarily studied.And it provided an idea for the research of measurement method of human blood pressure waveform.