Obstructive Sleep Apnea Detection Based On Spectral Analysis Of Heart Rate Without Resampling

Sleep apnea is a common sleep disorder, with a reported prevalence of 2 percent in women and 4 percent in men over the age of 35, and now is considered as an important risk factor for the development of cardiovascular diseases. There are three types of sleep apnea: obstructive, central and mixed. And Obstructive sleep apnea (OSA), of course, is the most common one. The traditional methods for assessment of sleep-related breathing disorders are sleep studies (polysomnography), with the recording of electro-encephalography (EEG), electro-oculography (EOG), electromyography (EMG), electrocardiography (ECG), oronasal airflow, respiratory effort and oxygen saturation. Sleep studies are expensive for patients, because they require overnight evaluation in sleep laborator-ies, with dedicated systems and attending personnel. So an effective and inexpensi-ve screening is highly desirable. The detection of the OSA using the ECG signal relies on the heart rate variability (HRV) that is related to the periodic cycles of breathing cessation and restoration. The effect of OSA on the PSD of the HR time series is to introduce frequency components in the frequency range of 0.01 Hz to 0.08 Hz. Traditional PSD estimation methods, such as periodogram and AR methods, operate on time series that is evenly sampled. Since HR time series is unevenly sampled, to apply these traditional techniques, one need to use interpolation to resample the HR time series. But the interpolation and resampling may introduce additional alteration of frequency content, especially when the HR time series is corrupted by noise (QRS detector errors or ectopic beats). In such situation, the performance of traditional PSD estimation methods can be severely affected. The classical periodogram and Lomb periodogram are methods for PSD estimation based directly on unevenly sampled time series, so completely avoid the problems related to resampling and sample replacement. This paper uses classical periodogram and Lomb periodogram with a windowing function to produce PSD estimates to detect the OSA, and then...