Frontiers in Applied Mathematics and Statistics (Apr 2021)
Tracking Rhythms Coherence From Polysomnographic Records: A Time-Frequency Approach
Abstract
The crosstalk between organs plays a crucial role in physiological processes. This coupling is a dynamical process, it must cope with a huge variety of rhythms with frequencies ranging from milliseconds to hours, days, seasons. The brain is a central hub for this crosstalk. During sleep, automatic rhythmic interrelations are enhanced and provide a direct insight into organ dysfunctions, however their origin remains a difficult issue, in particular in sleep disorders. In this study, we focus on EEG, ECG, and airflow recordings from polysomnography databases. Because these signals are non-stationary, non-linear, noisy, and span wide spectral ranges, a time-frequency analysis, based on wavelet transforms, is more appropriate to handle this complexity. We design a wavelet-based extraction method to identify the characteristic rhythms of these different signals, and their temporal variability. These new constructs are combined in pairs to compute their wavelet-based time-frequency complex coherence. These time-frequency coherence maps highlight the occurrence of a slowly modulated coherence pattern in the frequency range [0.01–0.06] Hz, which appears in both obstructive and central apnea. A preliminary exploration of a large database from the National Sleep Research Resource with respiration disorders, such as apnea provides some clues on its relation with autonomic cardio-respiratory coupling and brain rhythms. We also observe that during sleep apnea episodes (either obstructive or central), the cardiopulmonary coherence (in particular respiratory sinus-arrhythmia) in the frequency range [0.1–0.7] Hz strongly diminishes, suggesting a modification of this coupling. Finally, comparing time-averaged coherence with heart rate variability spectra in different apnea episodes, we discuss their common trait and their differences.
Keywords
