Characterizing cardiorespiratory interaction in preterm infants across sleep states using visibility graph analysis

Abstract

Cardiorespiratory interaction (CRI) has been intensively studied in adult sleep, yet not in preterm infants, in particular across different sleep states including wake (W), active sleep (AS), and quiet sleep (QS). The aim of this study was to quantify the interaction between cardiac and respiratory activities in different sleep states of preterm infants. The postmenstrual age (PMA) of preterm infants was also taken into consideration. The CRI during sleep was analyzed using a visibility graph (VG) method, enabling the nonlinear analysis of CRI in a complex network. For each sleep state, parameters quantifying various aspects of the CRI characteristics from constructed VG network including mean degree (Dm) and its variability (Dsd), clustering coefficient (CCm) and its variability (CCsd), assortativity coefficient (AC) and complexity (DSE) were extracted from the CRI networks. The interaction effect of sleep state and PMA was found to be statistically significant on all CRI parameters but AC and DSE. The main effect between sleep state and CRI parameters was statistically significant except for CCm, and that between PMA and CRI parameters was statistically significant but DSE. In conclusion, the CRI of preterm infants is associated with sleep states and PMA in general. For preterm infants with a larger PMA, CRI has a more clustered pattern during different sleep states, where QS shows a more regular, stratified, and stronger CRI than other states. In the future, these parameters can be potentially used to separate sleep states in preterm infants.

Characterising the motion and cardiorespiratory interaction of preterm infants can improve the classification of their sleep state

“…Several parameters of the visibility graph network topology constructed from the cardiorespiratory interaction time series exhibited significant differences among the preterm infant sleep states. 15 Therefore, we extracted eight features from the cardiorespiratory interaction signals. The means and standard deviations of the degree of the cardiorespiratory interaction network were constructed using the visibility graph method and the network created with the difference visibility graph 22 was calculated.…”

“…The assortative coefficient is a measure of assortative mixing by degree and it represents the skewness in the connections of network nodes. 15 The clustering coefficient quantified the density of the local clusters within a network. The means and standard deviations of the clustering coefficient values were calculated for each epoch.…”

“…We also extracted features from the degree distribution to examine the statistical properties of the cardiorespiratory interaction network. 15 An overview of the cardiorespiratory interaction network features is provided in Table S4.…”

“…Cardiorespiratory interaction (CRI) is the interaction between cardiac and respiratory dynamics. 15 For each 30-second epoch, this interaction is constructed by identifying the heartbeats and determining the amplitude of respiratory signal at the corresponding time stamp. Our previous studies demonstrated an association between cardiorespiratory interaction and preterm infant sleep state using the visibility graph method.…”

“…In this study, we proposed extracting motion signal features obtained via the motion detection method and from the cardiorespiratory interaction signals derived from the visibility graph complex network analysis. 15 The aim of this study was to improve the sleep classification accuracy in preterm infants by incorporating these novel features along with the ECG features proposed by Werth et al 19 and the respiratory features commonly used in adult sleep staging. 20…”