Apnea patients characterized by 0.02 Hz peak in the multitaper spectrogram of electrocardiogram signals

We present a method for identifying patients suffering from the sleep disorder apnea on the basis of their Electro-Cardiograms (EGCs). The method is a simple algorithm which utilises charateristic features present in the ECG spectrogram. Additionally we describe extensions which enable assessment of whether a subject is experiencing apnea on a minute by minute basis. We find that the most informative spectral feature is a previously observed low frequency oscillation at 0.02 Hz. The oscillation is present during apnea episodes in both the raw ECG signal and also the point process signal formed from the beat times. This indicates that there is a change in the intrinsic dynamics of the system regulating the cardiac cycle heart during apnea episodes The method was applied to a test set of data provided by Dr Penzel for the 'Computers in Cardiology Challange 2000' competition and classified all subjects correctly (30/30 = 100%) and scored 15257/17268 = 88% on the minute by minute classification.. Entrant 7 Entry numbers 20000503.123923 and 20000916.193931.