Study On Heart Function Changes Under Exercise Stress Based On Heart Sound

In recent years,with the continuous development and progress of the society,people pay more and more attention to the health problem,and exercise and fitness become more and more popular,while a series of problems caused by excessive exercise follow.Studies have shown that excessive exercise may lead to exercise-induced cardiac fatigue(EICF),myocardial injury,and heart disease or even sudden cardiac death in severe cases.Heart sound is the sound produced when the heart pumps blood periodically.It can detect the changes in heart function and structure earlier from the angle of myocardial force variability.Therefore,based on the relationship between heart sound and heart force,this paper carried out the exhaustive swimming experiment of New Zealand rabbits and the crowd weight-bearing cross-country experiment,extracted and analyzed the changes of cardiac reserve indexes in exercise stress experiment,and analyzed the changes of cardiac function under exercise stress based on machine learning.The main contents are as follows:Firstly,exercise stress experiment of New Zealand rabbits and exercise stress experiment of people were carried out.Based on the improved Viola integral method and double threshold method,cardiac reserve indexes were extracted,and the changes of cardiac reserve indexes of New Zealand rabbits and people were explored with the exercise stress experiment.The results showed that there were significant differences in cardiac reserve indexes between the surviving New Zealand rabbits and the sudden death New Zealand rabbits during exercise stress experiment.Immediately after exercise stress,the changes of cardiac reserve indexes in the population were also statistically significant,which further proved that cardiac reserve indexes could be used to evaluate cardiac function status.Secondly,based on convolutional neural network(CNN)and gated recurrent unit(GRU),a network for classification and recognition of heart sounds in New Zealand rabbits who survive and die suddenly under exercise stress was constructed.The CNN-GRU,CNN,GRU and extreme learning machine(ELM)were evaluated by 5-fold cross validation using heart sound signals of New Zealand rabbits 96 hours after exercise stress(survival sample)and before sudden death(sudden death sample)under exercise stress.The results showed that CNN-GRU had the best performance,with an average accuracy of 89.57%,sensitivity of 89.38% and specificity of 92.20%.In addition,the heart sounds of New Zealand rabbits at four different time points in the exercise stress experiment were input into CNN-GRU.It was found that with the progress of the experiment,the network’s ability to recognize the two kinds of heart sounds became stronger and stronger,especially in the second swimming exhaustion after 24 hours,the network could basically recognize the two kinds of signals,which proved the feasibility of deep learning in the exploration of sudden death caused by exercise.Thirdly,a study on the prediction of cardiac reserve index under crowd exercise stress based on machine learning was carried out.The mean absolute error,root mean square error and mean absolute percentage error were used to compare the prediction performance of ELM,BP neural network,CNN,GRU and CNN-GRU on cardiac reserve index by 10-fold cross validation.The results showed that compared with other networks,ELM showed better performance in predicting cardiac reserve indexes,which could be used to analyze cardiac function under exercise stress.