Time-variant spectral analysis of the heart rate variability during sleep in healthy and obstructive sleep apnoea subjects

A time-variant autoregressive approach was used in order to evaluate the spectral parameters of the heart rate variability (HRV), through the all sleep stages, in both normal subjects and patients with severe obstructive sleep apnoea. Recordings coming from five normal and five pathologic subjects were analyzed in the study. The parameters of the autoregressive model were fixed for the entire night recordings: model order = 8 and forgetting factor = 0.98. The classical spectral indexes of the heart rate variability were normalized respect to the total power. The results in this study showed, in normal subjects, an increment in the HFn components in NREM, while pathologic subjects presented a reduced activity in this component in all the time, which suggest a low activation of the vagal nerve. Furthermore, in both groups of subjects, VLFn reached high levels during REM and wake than NREM. In conclusion, this method could offer an alternative approach, with high resolution and efficient computation in the spectral decompotition, in order to develop a classification of sleep stage and apnoea detectors from the HRV.