Research On The Relationship Between Heart Sound Features And Respiratory Phase

The heart sound is a series of mechanical vibrations generated by the interaction of heart hemodynamics and the cardiovascular system.It carries a large amount of information about the health status of the cardiovascular system and is an important source of information for the diagnosis of heart disease and assessment of cardiac function.In the calm state of the human body,respiration is an important factor causing cardiac hemodynamic changes.Based on the physiological relationship chain of “respiration-cardiac hemodynamics-heart sound morphology”,this paper study the effect of respiration on heart sounds and establishes the model between respiratory phase and heart sound features.The work of this paper mainly includes the following two parts.(1)Study on the splitting of the second heart sound during respiration.The physiological mechanism of the cyclical changes in cardiac hemodynamics induced by respiration is studied.Based on the splitting relationship between the aortic valve component(A component)and the pulmonary artery component(P component)of the second heart sound,the second heart sound model under the respiratory state is proposed and a superposition and average method for quantitatively estimating the split value is proposed.Under simulation conditions,the performance of the proposed method,the STFT method and the Hilbert vibration decomposition method(HVD method)are compared.The simulation results show that the root mean square errors of the proposed method,HVD method and STFT method are 0.98 milliseconds,0.84 milliseconds,and 1.1 milliseconds,respectively.Further,the synchronized ECG,heart sound signals,and respiratory signals of 12 healthy adults are collected to verify the effectiveness of the methods.The above three methods are compared from the aspects of computational complexity,estimation of the splitting value,and the consistency of respiratory phases.The experimental results show that the method in this paper has the least complexity,the computation time is about a few hundredth of the HVD method,and it is about one tenth of the STFT method.The proposed method is slightly better than the HVD method or equivalent,higher than the STFT method in the degree of consistency about estimating splitting value.(2)The prediction of respiratory phases using heart sound features.The morphological changes of heart sounds caused by respiration are studied,and the features of heart sounds related to respiration are found.There are mainly the features of amplitude,energy,time domain and Intrinsic Mode Function(IMF)of heart sounds.The relationships between respiratory phase and heart sound features are modeled by RBF neural network,GRNN neural network and support vector machine.The models are tested with respiratory signals and heart sound signals from 22 healthy adults.The results show that there is a specific non-linear relationship between the respiratory phase and heart sound features,and the neural network can fit this nonlinear relationship.Through the experimental data,the prediction performance of each model is quantitatively evaluated.The analysis shows that the GRNN neural network has the best prediction performance and the respiratory phase prediction error is between 0.23 radians to 0.87 radians.The above research results are conducive to the quantitative evaluation of the physiological relationship chain of “respiration-cardiac hemodynamics-heart sound morphology”,forming a method of indirect assessment of cardiac function,and have potential applications in the screening of cardiac hemodynamic abnormalities and abnormal respiratory function.