| With the changes of lifestyle, the incidence of cardiovascular disease is increasingyear by year, and more and more young people caught this disease. According to thereports, cardiovascular disease is the top risk in the death list. Human physiological signalhas been used widely in clinical diagnosis of cardiovascular disease, which is the mosteffective means in clinical diagnosis of cardiovascular disease. The physiological signalsare complex in our body，including electrophysiological signals and non-electrophysiologicalsignals. The heart sound signal and pulse wave are the typical non-electrophysiologicalsignals. The two signals contain a large of information in the cardiovascular system. It canprovide a reliable base to detect and diagnose the cardiovascular diseases. And it is thefocal topic in the field of Biomedical Engineering and clinical research.This paper integrated and improved the previous analysis algorithms, and thenanalyzed the characteristics of the heart sound signal and pulse wave, extracted theeigenvalues of the two signals, and found the correlation parameters by SPSS factoranalysis, built a9-dimensional feature vectors[d1,d2,d3,d4,a4,ES1,ES2,Ti,Tj], where diis thecoefficient of wavelet decomposition, which represent the high-frequency component ini-layer; a4is the wavelet entropy; The ES1and ES2is the Shannon energy of the first heartsound and the second heart sound; Tiand Tjis the time interval between the first heartsound and the second heart sound. Combine with the information fusion technology,choose these factors as the input node of the BP neural network, at last the BP neuralnetwork can distinguish four types of mitral stenosis, aortic insufficiency, mitralregurgitation and aortic stenosis. A multi-parameter physiological information fusionmodel was built. This method can be used in the early detection of cardiovascular disease,which is a potential application in family healthcare, healthy self-test and early warningetc.The main contents were involved in this paper as following:1Analyze the characteristics of the heart sound signal and pulse wave. Using theimproved hardware system, the signals of heart sound signal and pulse wave werecollected. A visualize interface system was designed based on the Matlab GUI. The system can achieve the acquisition of the heart sound signal and pulse wave, andanalyze the factors in the early detection of cardiovascular diseases. It will provide anew way for the early diagnosis of cardiovascular diseases.2Comparing the filtering effect of EMD filter,0–phase filter and wavelet filter. thewavelet filter was selected to denoise the heart sound signal and pulse wave. TheShannon entropy was adopted to get the heart sound signal envelop and the position ofthe first heart sound and the second heart sound. At last, using the HHT, the values oftime and frequency was analyzed.3Extract the characteristic parameters, including the wavelet entropy of each layer,Shannon energy and time intervals etc. using the wavelet analysis, the Shannon entropyand time limit. The impact factor analysis was performed with SPSS to remain themost valuable features and remove the irrelevant features.4All the characteristic values were combined into a matrix containing nine elements,which was regarded as the input of a BP neural network for the identification of heartsound signals. The recognition rate of the simple aortic regurgitation, the aorticregurgitation, the mitral valve stenosis and mitral valve insufficiency were73.33%,80.00%,86.67%and93.33%, respectively. In order to improve the recognition rate ofBP neural network, the variable leaning rate and crossover study was adopted. Theresults show that the recognition rate for the patient of cardiovascular disease reaches88.33%, and the recognition accuracy for the healthy person is up to91.67%. |
PDF Full Text Request