Computer-assisted analysis of pleural and subpleural lung ultrasound correlates with oxygenation in preterm infants

Lung ultrasound (LUS) is increasingly used in neonatal intensive care, however visual scoring of LUS images remains subjective. Quantitative LUS (Q-LUS) is objective and whilst it detects small lung aeration changes in preterm animals, human data are limited. We hypothesised that Q-LUS would correlate strongly with oxygenation in very preterm infants born before 32 weeks’ gestation. We examined two quantitative image properties to assess whether Q-LUS correlates with oxygen status in very preterm infants: (1) mean-grey-value (MGV) and (2) second-order textural features (angular second moment, contrast, grey-level correlation (Q-LUScorrelation), inverse difference moment, entropy). 560 LUS images were acquired from 70 very preterm infants using Venue 50 and Venue Go systems and analysed in Fiji ImageJ. Oxygenation was assessed using oxygen saturation index (OSI) and peripheral oxygen saturation to fraction of inspired oxygen (S/F) ratio. Spearman’s rank correlation coefficient (ρ) was used to determine correlations between Q-LUS measurements and infant oxygenation. We found that MGV correlated fairly with OSI (ρ=-0.46, [-0.68,-0.23], p