716 - Studying preterm respiration in the Neonatal ICU with the Analysis of Respiratory Kinematics (ARK) System

Publication Number: 4700.716

William Ashe, University of Virginia School of Medicine, Charlottesville, VA, United States; Brynne A. Sullivan, University of Virginia, CHARLOTTESVILLE, VA, United States; Shrirang M.. Gadrey, University of Virginia School of Medicine, Charlottesville, VA, United States; Angela Gummadi, University of Virginia School of Medicine, Charlottesville, VA, United States

Background: Premature infants often require respiratory support due to underdeveloped lungs, immature breathing control, and weak muscles, yet clinical decisions rely on subjective assessments and standard NICU technology fails to capture clinically relevant patterns, such as increased effort. The Analysis of Respiratory Kinematics (ARK) system, developed using non-invasive inertial measurement sensors and motion signal algorithms, detects labored breathing in hospitalized adults and has not been tested in neonates.

Objective: To assess the technical feasibility and clinical utility of respiratory kinematic measurement in preterm infants using ARK technology.

Design/Methods: In preterm infants admitted to the NICU, we recorded 30 minutes of chest and abdominal motion using a 2-sensor ARK system, adapted from the 7-sensor system validated in adults. We divided each recording into 2-minute segments and applied an acceptability criterion of ≥30 seconds of artifact-free signal. After filtering and processing, we extracted 29 clinically meaningful kinematic features, including a Rib-Abdomen Dominance (RAD) index calculated from the ratio between compartmental amplitudes (Figure 1). We used two-sample t-tests to compare kinematics between neonates with and without significant respiratory support (CPAP or HFNC), applying Bonferroni correction for 29 comparisons. For features differing by support status, we assessed correlations with postmenstrual age (PMA), birth weight (BW), and sex.

Results: We enrolled 36 infants with a median gestational age (GA) of 32.1 weeks and median PMA at time of ARK recording of 33.9 weeks, obtaining 60 recordings (Table 1). During 23 ARK recordings, the neonate was on respiratory support, providing positive pressure for respiratory distress (CPAP or HFNC), while the remaining 37 (62%) were without support. Of the 29 comparisons, the RAD index was the most discriminative between support groups (Figure 1); neonates requiring significant respiratory support exhibited rib-dominant breathing (positive RAD), while those without support had strong abdominal dominance (negative RAD; RAD index +0.29 vs -0.32; corrected p < 0.05). The RAD Index correlated with PMA and BW, with abdominal dominance reflecting greater maturity. No sex differences were observed. Exploratory visualizations revealed patterns of clinical interest, such as apnea (Figure 2).

Conclusion(s): Respiratory kinematic features were technically feasible to acquire in preterm neonates and demonstrated meaningful associations with indicators of neuromuscular maturation and dependence on respiratory support.

Table 1. Cohort characteristics


Figure 1: Rib-Abdomen Dominance (RAD) Index
Panel A shows the kinematics of a neonate that was not requiring any respiratory support [Postmenstrual age (PMA) = 33.6 weeks; RR = 53 bpm]. Panel B shows the kinematics of a neonate requiring CPAP [PMA = 32.4 weeks; RR = 58 bpm]. [For visual clarity, only one accelerometer axis is shown in Panels A-B; the RAD Index is derived from 3-axis accelerometer data.] Despite similar PMA and RR, the infants in panels A & B exhibit a markedly different RAD Index. In the neonate requiring CPAP, the rib amplitude exceeds abdominal amplitude, yielding a positive RAD Index (+1.9). In the neonate not requiring any support, the opposite pattern (i.e., abdominal dominance) is displayed, resulting in a negative RAD Index (-2.1). Panel C shows a higher distribution of RAD index measurements among preterm infants on respiratory support compared to those not on respiratory support (p < 0.05 after Bonferroni correction for multiple comparisons). Panel D shows the RAD index plotted as a function of PMA with a negative correlation, indicating more rib-dominant breathing kinematics at lower PMA.

Figure 2: Exploratory visualization of apnea and periodic breathing
These are representative examples of regular (panel A) and periodic (panel B) breathing patterns observed in exploratory kinematic visualizations. In the periodic breathing example, there are cyclical increases and decreases in signal amplitude during periods of rapid breathing followed by short apneas. In contrast, the example of regular breathing contains neither interval nor amplitude variations. Both of these patterns vary from the signals in Figure 1, where breathing exhibits amplitude modulation without pauses. These examples demonstrate the rich physiological information encoded in thoraco-abdominal kinematics-information not captured by respiratory rate alone. [For visual clarity, only one accelerometer axis is shown; quantitative analyses are performed on 3-axis data.]