Estimation of respiratory cycle by non-uniform sampling heart rate variability time series

Abstract

It has been demonstrated that sleep plays an important role for preventing dementia and other types of other diseases. Therefore, the sleep evaluation is important to find sleep-related problems and to improve sleep quality. In this study, to achieve an accurate and easy evaluation of the sleep quality, we developed a sleep stages classification method based on heart rate variability (HRV). Here, we focus on respiratory sinus arrhythmia (RSA) and propose a novel method to accurately detect the cycle length of periodic breathing. Our method analyzes periodically averaged HRV patten and can estimate the breathing cycle with higher resolution than that of conventional Fourier-transform-based methods. We applied our method to HRV during sleep and calculate RSA-related periodicity and amplitude parameters. Using these features together with conventional HRV parameters, we trained random forest to classify REM and non-REM sleep stages. Our result demonstrated that RSA-related features can improve the classification accuracy.