Research On Predicting Method Of Cardiovascular Disease By Multimodality Information Fusion

Nowadays,cardiovascular disease has posed a great threat to human life and health,through analyzing and processing of cardiovascular disease data,the cardiovascular disease can be predicted and diagnosed timely.Therefore,the thesis takes the cardiovascular disease prediction as the research object,the multimodal information fusion technology of forecasting the cardiovascular disease as the focus in research,to study the method of predicting cardiovascular disease.The main works include the following several aspects:(1)Analyzing and selecting the diagnosis of factors which affect the cardiovascular disease.Summed up all kinds of medical check data and data types that is required to diagnosis the cardiovascular disease.Due to the limitation of the data source,this thesis selects the electrocardiogram,heart sounds,and digital data type medical data to analyze and process.(2)Converting each model's medical data into a digital type,the electrocardiogram uses the MIT / BIH standard ECG database as a remote data and test standards,and uses the Mallat algorithm to calculate the R-wave peak and the starting and ending point of the QRS wave,finally the feature points' time and amplitude characteristics are used as reference for pathological diagnosis.(3)For the characteristics and the data type of heart sound signal,taking use of wavelet de-noising,using the db6 wavelet to divide the frequency band,the Shannon energy is used to extract the envelope,and the position of the beginning and end of the heart sound is expressed by the time limit,so as to complete the detection of the heartsound characteristic.(4)Taking heart disease as an example,thesis used the PCA to reduce the heart disease data set of Cleveland Fund Clinic,and select the main component.Combined with the expert scoring,improved PCA algorithm to calculate the weight of principal component.(5)The thesis makes use of SVM to classify a number of physiological information processed by PCA.The Guass Radial Basis Function is selected and select the kernel function parameter ?,as well as the penalty factor C,choose the optimal predicted results.Finally,compared with the predictions of the BP neural network and the classical SVM.