Predicting the termination of paroxysmal atrial fibrillation (AF) may provide a signal to decide whether there is a need to intervene the AF timely. We proposed a novel RdR RR intervals scatter plot in our study. The abscissa of the RdR scatter plot was set to RR intervals and the ordinate was set as the difference between successive RR intervals. The RdR scatter plot includes information of RR intervals and difference between successive RR intervals, which captures more heart rate variability (HRV) information. By RdR scatter plot analysis of one minute RR intervals for 50 segments with non-terminating AF and immediately terminating AF, it was found that the points in RdR scatter plot of non-terminating AF were more decentralized than the ones of immediately terminating AF. By dividing the RdR scatter plot into uniform grids and counting the number of non-empty grids, non-terminating AF and immediately terminating AF segments were differentiated. By utilizing 49 RR intervals, for 20 segments of learning set, 17 segments were correctly detected, and for 30 segments of test set, 20 segments were detected. While utilizing 66 RR intervals, for 18 segments of learning set, 16 segments were correctly detected, and for 28 segments of test set, 20 segments were detected. The results demonstrated that during the last one minute before the termination of paroxysmal AF, the variance of the RR intervals and the difference of the neighboring two RR intervals became smaller. The termination of paroxysmal AF could be successfully predicted by utilizing the RdR scatter plot, while the predicting accuracy should be further improved.
Sleep apnea (SA) detection method based on traditional machine learning needs a lot of efforts in feature engineering and classifier design. We constructed a one-dimensional convolutional neural network (CNN) model, which consists in four convolution layers, four pooling layers, two full connection layers and one classification layer. The automatic feature extraction and classification were realized by the structure of the proposed CNN model. The model was verified by the whole night single-channel sleep electrocardiogram (ECG) signals of 70 subjects from the Apnea-ECG dataset. Our results showed that the accuracy of per-segment SA detection was ranged from 80.1% to 88.0%, using the input signals of single-channel ECG signal, RR interval (RRI) sequence, R peak sequence and RRI sequence + R peak sequence respectively. These results indicated that the proposed CNN model was effective and can automatically extract and classify features from the original single-channel ECG signal or its derived signal RRI and R peak sequence. When the input signals were RRI sequence + R peak sequence, the CNN model achieved the best performance. The accuracy, sensitivity and specificity of per-segment SA detection were 88.0%, 85.1% and 89.9%, respectively. And the accuracy of per-recording SA diagnosis was 100%. These findings indicated that the proposed method can effectively improve the accuracy and robustness of SA detection and outperform the methods reported in recent years. The proposed CNN model can be applied to portable screening diagnosis equipment for SA with remote server.