Analysis of ventilatory pattern variability

Abstract

Each breath is not generated de novo; rather, the ventilatory pattern is a continuous oscillation in which the next breath is related to the present one; and being biologic, the ventilatory pattern varies. Further, the responsiveness of respiration to sensory input is dynamic because neural mechanisms scale afferent input. Thus, ventilatory pattern variability (VPV) has deterministic properties, which may vary in health and disease. We have developed analytical tools to distinguish and assess linear and nonlinear sources of VPV. Surrogate data sets obtained by shuffling the original data while preserving its amplitude distribution and autocorrelation function and, thus, preserving linear properties embedded within the original data are used to distinguish various sources and types of VPV. Differences in mutual information and sample entropy of VPV between original and surrogate data sets reflect nonlinear deterministic properties of the original data set. We have applied these analytic techniques to assess breathing pattern before and after vagotomy, cerebral ischemia, and lung injury. Deterministic variability decreased following each of these interventions. Finally, our approach can be applied to rhythmic biological signals.