Cerebral oxygen saturation and peripheral perfusion in the extremely premature infant with intraventricular and/or pulmonary haemorrhage early in life

Abstract

Extremely preterm infants are at higher risk of pulmonary (PH) and intraventricular (IVH) haemorrhage during the transitioning physiology due to immature cardiovascular system. Monitoring of haemodynamics can detect early abnormal circulation that may lead to these complications. We described time-frequency relationships between near infrared spectroscopy (NIRS) cerebral regional haemoglobin oxygen saturation (CrSO₂) and preductal peripheral perfusion index (PI), capillary oxygen saturation (SpO₂) and heart rate (HR) in extremely preterm infants in the first 72 h of life. Patients were sub-grouped in infants with PH and/or IVH (N _H = 8) and healthy controls (N _C = 11). Data were decomposed in wavelets allowing the analysis of localized variations of power. This approach allowed to quantify the percentage of time of significant cross-correlation, semblance, gain (transfer function) and coherence between signals. Ultra-low frequencies (<0.28 mHz) were analyzed as slow and prolonged periods of impaired circulation are considered more detrimental than transient fluctuations. Cross-correlation between CrSO₂ and oximetry (PI, SpO₂ and HR) as well as in-phase semblance and gain between CrSO₂ and HR were significantly lower while anti-phase semblance between CrSO₂ and HR was significantly higher in PH-IVH infants compared to controls. These differences may reflect haemodynamic instability associated with cerebrovascular autoregulation and hemorrhagic complications observed during the transitioning physiology.

Figure 1

Boxplots of (a) near infrared spectroscopy (NIRS) cerebral regional haemoglobin oxygen saturation (CrSO₂), (b) peripheral perfusion index (PI), (c) heart rate (HR) and (d) capillary oxygen saturation (SpO₂) in PH-IVH patients (left boxplots) and healthy age-matched controls (right boxplots). On each box, the central mark is the median, the square is the mean, the stars are the individual data, the edges of the box are the 25^th and 75^th percentiles, and the whiskers show the 95% confidence interval. Empty circles denote outliers and statistical comparisons are indicated with corresponding p-values (n.s., non-significant).

Figure 2

Temporal distributions of (a) near infrared spectroscopy (NIRS) cerebral regional haemoglobin oxygen saturation (CrSO₂), (b) peripheral perfusion index (PI), (c) heart rate (HR) and (d) capillary oxygen saturation (SpO₂) in PH-IVH patients (left column) and healthy age-matched controls (right column) in the first 72 h of life. On each plot, the black curve is the mean and the grey shaded region represents one standard deviation of the group.

Figure 3

Example of the complete analytical workflow in a healthy control (left column) and in an infant with pulmonary (PH) and/or intraventricular haemorrhage (IVH, right column): (a) and (b) depict temporal distributions of near infrared spectroscopy (NIRS) cerebral regional haemoglobin oxygen saturation (CrSO₂) and peripheral perfusion index (PI) in the first 72 h of life, respectively; (c–f) display the amplitude of the cross-correlation, the semblance (anti-phase and in-phase), the amplitude of the gain (transfer function) and the coherence between CrSO₂ and PI in the time-frequency space. Regions that are statistically significant are comprised in a black bold contour. A dashed white line indicates the selected ultra-frequency band of slow and prolonged periods of >1 h (<0.28 mHz) used for statistical analysis. Regions outside the cone of influence in which data are not used in the statistical analysis below are not shown.