Estimation Of Left Ventricle Blood Pressure Parameters Based On External Heart Sounds Of Beagle Dogs

Heart sounds are mechanical vibrations which are produced by the interplay between cardiac haemodynamics and valves,heart chambers,surrounding tissues.Any change of cardiac haemodynamics can lead to varying heart sound features.Hence,there are natural relationships between the heart sound features and the cardiac haemodynamics.On the basis of animal experiments,heart sound features were studied in this paper.Selected features were used in estimation of the left ventricle systolic pressure,the diastolic pressure and the maximum rising rate of blood pressure.This study would provide a technological approach to monitor cardiac haemodynamics noninvasively and continuously.In this thesis,the main work is focused on the following two aspects.(1)On the basis of animal experiments,the relationships between heart sound features and left ventricle blood pressure are investigated.Three beagle dogs were used as experimental animals,and their electrocardiograms,heart sounds and left ventricle blood pressure signals were collected synchronously.37 data records were obtained in the animal experiments and the data records were analyzed.The dogs were injected with different doses of epinephrine to change their cardiac haemodynamics.The collected data records were preprocessed and processed by the operation of heart sound segmentation.Outliers were removed by the standardized residual and the leverage ratio.Heart sound features were defined and separated,which include amplitude features,frequency features,energy features,time features and entropy features.The correlations between heart sound features and the left ventricle systolic pressure,the diastolic pressure,the maximum rising rate of blood pressure were analyzed.According to the results,the left ventricle systolic pressure has a high correlation with total left ventricular systole and the first heart sound spectral width.Correlation coefficients are 0.89 and 0.84,respectively.The left ventricle diastolic pressure has a high correlation with total left ventricular systole and the first heart sound fuzzy entropy.Correlation coefficients are 0.71 and 0.67,respectively.The maximum rising rate of blood pressure has a high correlation with total left ventricular systole and the first heart sound spectral width.Correlation coefficients are 0.91 and 0.85,respectively.(2)On the basis of animal experiments,estimation of the left ventricle systolic pressure,the diastolic pressure and the maximum rising rate of blood pressure using heart sound features is investigated.Several models were involved in estimation of left ventricle blood pressure parameters.They were multi-parameter linear regression model,decision tree model,neural network model and support vector machine model.Those heart sound features having high correlation with the target are selected as model input.Differences between measured values and estimated values were compared to evaluate each model.The results indicate that on the basis of heart sounds,the estimation of left ventricle systolic pressure,diastolic pressure and maximum rising rate of blood pressure is viable.Support vector machine model performs well in estimation of the left ventricle systolic pressure and the diastolic pressure.The average estimation erroes are 6.16 mmHg and 1.79 mmHg.Decision tree model performs well in estimation of the left ventricle maximum rising rate of blood pressure.The average estimation error is 0.33 mmHg/ms.The results of this study provide a technical approach for noninvasive monitor of cardiac hemodynamics continuously.