Objective: In this study, we propose a method to detect Obstructive sleep apnea (OSA) based on electrocardiogram (ECG) signals based on machine learning using Support Vector Machine (SVM) algorithm.

Background: Obstructive sleep apnea (OSA) is a common form of the disorder that causes sufferers to experience respiratory arrest during sleep, which can lead to cardiovascular disease if not treated properly. OSA is commonly found in patients with Parkinson’s disease (PD). In general, apnea testing usually uses polysomnography (PSG), which is the standard procedure for the diagnosis of all sleep disorders. However, most cases of sleep apnea are currently undiagnosed due to the cost.

Method: Data in the form of ECG signals were obtained from Physionet Database using 12 different recording subjects. There are 70 ECG signal recordings recorded from different subjects with an age variation of 27-53 years old, both male and female. The data were recorded for two consecutive nights using single-channel ECG recording. The signals were then subjected to statistical feature extraction based on the distance between R peaks or RR intervals of ECG signals processed in a short duration period. The results of the feature extraction were used as input for classification using the Support Vector Machine method to be trained and tested on apnea and non-apnea recordings from OSA-positive and negative subjects. Tests were also conducted to determine its performance by obtaining the accuracy, sensitivity, and specificity values of the system.

Results: RR interval using 11 statistical features namely mean, standard deviation, median, two NN50 values (the number of pairs of successive NN (R-R) intervals that differ by more than 50 mS), two pNN50 values (the proportion of NN50 divided by the total number of NN (R-R) intervals.), the standard deviation of the NN (R-R) intervals (SDSD), root mean square of the successive differences (RMSSD), interquartile range, and mean absolute deviation (MAD) can work well, as evidenced by the system performance results achieving 83.7% accuracy, 79.4% sensitivity, and 87.4% specificity.

Conclusion: Based on the accuracy results obtained quite high exceeding 80%, the system can be said to be quite successful in carrying out its detection. However, further research is needed to obtain better results.

« Back to 2023 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/classification-of-obstructive-sleep-apnea-based-on-statistical-features-of-the-rr-interval-using-the-support-vector-machine-algorithm/