Application of TQWT based filter-bank for sleep apnea screening using ECG signals

The sleep apnea is a disease in which there is the absence of airflow during respiration for at least 10 s. It may occur several times during the night sleep. This disease can lead to many types of cardiovascular diseases. To detect this disease, signals obtained from many channels of polysomnography are to be observed visually by physicians for the long duration. This procedure is expensive, time-consuming, and subjective. Hence, it is required to build an automated system to detect the sleep apnea with few channels. This paper uses single-lead electrocardiogram (ECG) signal to detect apneic and non-apneic events. The proposed method uses tunable-Q wavelet transform (TQWT) based filter-bank instead of TQWT to decompose the segment of ECG signal into several constant bandwidth sub-band signals. Then centered correntropies are computed from the various sub-band signals. The obtained features are then fed to the various classifiers to select the optimum performing classifier. In this work, we have obtained the highest classification accuracy, specificity, and sensitivity of 92.78%, 93.91%, and 90.95% respectively using random forest classifier. Hence, our developed prototype is ready for validation with the huge database and clinical usage.