Heart sound is one of the common medical signals for diagnosing cardiovascular diseases. This paper studies the binary classification between normal or abnormal heart sounds, and proposes a heart sound classification algorithm based on the joint decision of extreme gradient boosting (XGBoost) and deep neural network, achieving a further improvement in feature extraction and model accuracy. First, the preprocessed heart sound recordings are segmented into four status, and five categories of features are extracted from the signals based on segmentation. The first four categories of features are sieved through recursive feature elimination, which is used as the input of the XGBoost classifier. The last category is the Mel-frequency cepstral coefficient (MFCC), which is used as the input of long short-term memory network (LSTM). Considering the imbalance of the data set, these two classifiers are both improved with weights. Finally, the heterogeneous integrated decision method is adopted to obtain the prediction. The algorithm was applied to the open heart sound database of the PhysioNet Computing in Cardiology(CINC) Challenge in 2016 on the PhysioNet website, to test the sensitivity, specificity, modified accuracy and F score. The results were 93%, 89.4%, 91.2% and 91.3% respectively. Compared with the results of machine learning, convolutional neural networks (CNN) and other methods used by other researchers, the accuracy and sensibility have been obviously improved, which proves that the method in this paper could effectively improve the accuracy of heart sound signal classification, and has great potential in the clinical auxiliary diagnosis application of some cardiovascular diseases.
The peak period of cardiovascular disease (CVD) is around the time of awakening in the morning, which may be related to the surge of sympathetic activity at the end of nocturnal sleep. This paper chose 140 participants as study object, 70 of which had occurred CVD events while the rest hadn’t during a two-year follow-up period. A two-layer model was proposed to investigate whether hypnopompic heart rate variability (HRV) was informative to distinguish these two types of participants. In the proposed model, the extreme gradient boosting algorithm (XGBoost) was used to construct a classifier in the first layer. By evaluating the feature importance of the classifier, those features with larger importance were fed into the second layer to construct the final classifier. Three machine learning algorithms, i.e., XGBoost, random forest and support vector machine were employed and compared in the second layer to find out which one can achieve the highest performance. The results showed that, with the analysis of hypnopompic HRV, the XGBoost+XGBoost model achieved the best performance with an accuracy of 84.3%. Compared with conventional time-domain and frequency-domain features, those features derived from nonlinear dynamic analysis were more important to the model. Especially, modified permutation entropy at scale 1 and sample entropy at scale 3 were relatively important. This study might have significance for the prevention and diagnosis of CVD, as well as for the design of CVD-risk assessment system.