Algorithm Research For Cardiac Reserve Indices Automatic Extraction

Cardiovascular disease threat human beings life and health seriously,has become a major public health problem.Heart sound(HS)is an important physiological signal,contains various physiological and pathological information of blood vessels,valve tissue organs associated with heart disease,whose detection and analysis is becoming one of important way for heart disease clinical auxiliary diagnosis.Cardiac reserve(CR)refers to one's ability to improve the heart function,which can evaluate the heart function state.Two main indicators for evaluating CR function in HS are the ratio of the first heart sound amplitude to second heart sound amplitude(S1/S2),and the ratio of diastolic to systolic duration(D/S).Referencing ECG signal is the traditional method for CR indices automatic extraction,with the continuous development of digital signal processing technology,HS characteristic envelope has become a primary way to extract CR indices.Research a rapid,effective and fully reflect time domain feature of HS envelope extracting algorithm is of great significance to clinical assistant diagnosis for cardiovascular disease.Firstly,the generation mechanism of HS and basic knowledge of CR were introduced in this thesis,which is good for understanding the relationship between HS and CR.And then analyze noises that may be mixed in HS when acquiring signal and many denoising methods were introduced.Design three groups of experiments to determine the parameters of wavelet denoising.Wavelet transform was adopted to realize HS denoising pretreatment.Traditional HS segment method based on a reference signal is high accuracy,but it'll make pretreatment become more complicated.HS segmentation using the characteristic envelope would be more simple and rapid.In this thesis three methods were researched,Hilbert huang transform,normalized Shannon energy and Viola integral wave method which are commonly used in HS characteristic envelope extraction.Compared with the characteristics of each method and application,a modified Viola integral method to extract the characteristic envelope of HS based on the advantages of Shannon energy has been put forward,which can extract characteristic envelope effectively,with strong anti-jamming and fast operation speed.Secondly,this thesis uses the single threshold method to realize HS segmention,combining with the short-time zero crossing ratio,an improved threshold segmentation algorithm is proposed.Combined with the feature of HS in the time domain,segment the HS further and automatic extract the CR indices using this modified method.Finally,in order to analysis of the human beings' heart function easily,the graphical interface has been designed to extract the CR indices,which is with complete functions,simple operation and processing fast,meet the expected requirement.In this thesis,CR indices of 317 cases of normal and abnormal HS had been extracted by using the modified Viola integral waveform method and threshold segmentation method.The accuracy is up to 99.87% and 90.86% respectively,in normal and abnormal HS signals.The result of HS statistical analysis shows that CR indices have reference value in HS classification.