Analysis And Identification Of Post-exercise Cardiac State Based On Heart Sound Features Combined With CTnI Detection

Several studies have shown that exercise has different effects on the heart.Heart sound can evaluate cardiac physiological state due to its direct reflection of the mechanical properties of cardiac activity.Cardiac injury biomarkers can effectively reflect the situation of cardiac injury.In this study,the rabbit exhaustion swimming experiment and the crowd intermittent weight-bearing cross-country running experiment were designed,and the cardiac physiological change after different intensity of exercise were evaluated based on heart sound feature extraction combined with cardiac troponin I(c Tn I)and myoglobin(Myo)detection.And the heart sound signals are classified by machine learning to realize the recognition of cardica state after exercise.The main contents are as follows:Firstly,study on heart sound signal processing method.An adaptive wavelet denoising algorithm was used to denoise heart sound signals.Then,heart rate(HR)and the ratio of diastolic to systolic duration(D/S)were solved based on signal envelope extraction and segmenting localization.Seven multifractal features parameters related to generalized Hurst exponent,multifractal spectrum and Renyi index were extracted by the multifractal detrended fluctuation analysis(MF-DFA).Secondly,the changes of cardiac physiology after exhaustion exercise were studied based on animal experiments.The rabbit exhaustion load-bearing swimming experiment was carried out,and HR and D/S values were extracted by heart sound signal processing,and serum c Tn I and Myo levels were detected before and after exercise.According to the survival and death status of the sample after exhaustion swimming experiment,the samples were divided into survival and death groups.Statistical analysis results show that irreversible myocardial injury occur after exhaustion swimming exercise in the death group,while in the survival group,exercise-induced cardiac fatigue occur immediately after exhaustion swimming exercise and recovered to normal after rest.Moreover,many times exhaustion swimming experiments had a tendency to enhance the cardiac reserve capacity of the surviving individuals.Thirdly,cardiac state changes have been studied based on human experiments after high intensity exercise.An experimental scheme of the crowd intermittent weight-bearing cross-country running was designed.Two time domain features and seven multifractal feature parameters were extracted from heart sound signals before and after exercise.Serum c Tn I content was detected before and after exercise.The results show that the compensation of motor cardiac function occurred immediately after 3 days of high-intensity exercise.After rest,HR was significantly decreased,and D/S was slightly increased,and multifractal intensity and complexity of heart sound signals are enhanced after 3-day cross-country running.These results indicate that the cardiac reserve capacity has a tendency to increase after the experiment of intermittent weight-bearing cross-country running.Fourthly,recognition of cardiac physiological state after exercise based on machine learning.The heart sound signals before and immediately after exercise were classified and recognized by three machine learning methods including support vector machine(SVM)、eatremely learning machine(ELM)and kernel extreme learning machine(KELM).The kernel function of SVM and KELM,as well as activation function of ELM were determined based on the accuracy.The parameters of three machine learning methods were selected by grid search algorithm.The results show that the accuracy,precision,recall and F1 values of KELM classification model are the highest,reaching93.94%,98.21%,88.71% and 93.22% respectively,and its training time is the shortest,so KELM is most suitable for classification and recognition of heart sound signals before and after exercise,so as to realize recognition of cardiac physiological state after exercise.