Research On Heart Failure Diagnosis Model Based On Heart Rate Variability

With the development of the social economy and the increment of aging of the population,the number of patient with cardiovascular diseases has increased dramatically and already reached 290 million until now.The mortality rate of cardiovascular disease is significantly higher than that of other diseases and has become the number-one killer of human health.In addition,the medical cost related to cardiovascular disease is up to 130 billion RMB.Heart failure,as one of the important categories of cardiovascular disease,has become a hot point in biomedical research.It will be of great significance to establish the early diagnosis model of heart failure disease and turn the traditional cure mode into the prevention model with the help of the development of computing technology in the background of the transformation of medical model in China.HRV is an effective diagnostic indicator of cardiovascular disease,however there are still amount of problems concerning the clinical diagnosis of heart failure should be further studied in the future.The aim of this study is to establish a diagnostic model of heart failure based on the in-depth study of the HRV analysis method to provide a clinical basis for the early diagnosis of heart failure.The in-depth study was conducted pointing at the HRV stability.HRV index validity and heart failure diagnosis model based on HRV.The RR interval of 54 healthy people and 29 heart failure patients in the PhysioBank database were used and 9 indexes of HRV including time domain index(MEAN,SDNN,RMSSD),frequency domain index(LFn,HFn,LF/HF)and nonlinear domain index(VAI,VLI,SampEn)were selected to conduct the analysis.The main research content is as follows:(1)The study on the stability of HRV index.The variance analysis and intra group correlation coefficient were selected as the research method and the Number Twenty was selected as the number of individual data segments by evaluating the impact of each individual data segment on the stability to analyze the stability of HRV in the heart failure population and the healthy population,respectively.The results showed that for healthy people,MEAN,LF/HF.LFn and HFn showed stablility,and for patients with heart failure.MEAN,VLI,SDNN,VAI,SampEn,LFn,HFn showed stablility.The stability of patients with heart failure is higher than that of healthy people.Recommendations for the stability of HRV were given in this paper.(2)The study on the effectiveness of HRV Indexes.The T-test was performed on 9 HRV indicators.The results showed that:HRV indicators have the ability to differentiate heart failure,and the discriminative ability from the largest to the smallest:VLI,SDNN,LFn&HFn&LF/HF,VAI,SampEn,RMSSD.(3)The study on the diagnosis model of heart failure based on HRV index.The diagnostic model for heart failure selected the grid search as a super parameter search algorithm,RBF-S VM and random forest as the classification model and 9 HRV indexes as input feature vectors.In addition,the framework nesting 10-fold cross-validation was used to establish the heart failure diagnosis model.The classification effects of random forest model were as follow:Sensitivity is 82.81%,Specificity is 87.06%,and Accuracy is 84.94%.The classification effects of RBF-SVM were as follow:Sensitivity is 80.74%,Specificity is 84.94%and Accuracy is 84.85%.By setting the cost of misclassification of heart failure to 1000,the sensitivity of the RBF-SVM model can be increased to 98.19%.The study provides a useful reference for machine learning in heart failure research.