Diagnosis of obstructive sleep apnea based on variational cardiopulmonary coupling

The electrocardiogram-based cardiopulmonary coupling (CPC) is widely applied for sleep-disordered breathing (SDB). However, the existing quantifications suffer spurious coupling, often arising from spectral leakage in Fourier-based CPC, or mode mixing and end effects with (ensemble) empirical mode decomposition (EMD/EEMD), which spreads respiratory sinus arrhythmia energy spectrally, thus obscuring intrinsic physiological dynamics. This article proposed variational cardiopulmonary coupling (VMD-CPC) to measure CPC dynamics in sleep. To validate our method, simulated data were crafted. Then, correlations between CPC features and apnea/hypopnea annotations, and also group comparisons (normal, mild & moderate, and severe) in features are conducted. The VMD-CPC maintains HF-normalized coupling power to the theoretical value of 1.0, outperforming that of Fourier-based CPC, EMD-CPC, and EEMD-CPC. The CPC spectrogram of sleep-disordered patients is featured by low-frequency coupling, while the dominant frequency of healthy individuals switches between low- and high-frequencies. Three indices, LF/HF, LFnorm, and HFnorm, show strong correlations with the apnea/hypopnea index (AHI), wherein LF/HF achieves r = 0.8449, presenting significant differences between the normalities and groups in different severities ( p < 0.0001). In all, our VMD-CPC provides a useful tool for SDB diagnosis, reducing spectral leakage and mode mixing that induce spurious coupling, offering signs to evaluate therapeutic effects.