Computer assisted analysis of the chest radiograph lung area and prediction of failure of extubation from mechanical ventilation in preterm neonates.

Post-extubation chest radiographs (CXRs) are frequently requested on the neonatal intensive care unit, but it is controversial whether they generate useful information. A low lung volume assessed by measurement of functional residual capacity (FRC) post extubation has been demonstrated to predict extubation failure, which is a subsequent requirement for increased respiratory support. We have previously shown that the CXR lung area obtained by computer assisted analysis significantly correlated with FRC and, therefore, speculated that a low CXR lung area post extubation would reliably predict extubation failure. The aim of this study was to test the hypothesis by analysing CXRs from 20 infants, with median gestational age of 28 weeks (range 25-33 weeks) and postnatal age 4 days (range 1-11 days). CXRs were obtained within 4 h of extubation and were scanned and analysed using a Power Macintosh computer with a Wacom A5 Ultra pad and NIH image software. The cardiac, mediastinal and thymic shadows, and areas of perihilar and lobar consolidation were subtracted from the thoracic area to give the lung area. Seven infants failed extubation and differed significantly from the rest of the cohort only with regard to their CXR lung area, median gestational age, birth weight and postnatal age. Receiver operator characteristic (ROC) curves were constructed and the areas under each ROC curve were compared. Analysis demonstrated that a low CXR lung area and an older postnatal age were the best predictors of extubation failure. A post-extubation CXR lung area of < 8.5 cm2 had the highest specificity (100%) in predicting extubation failure. We conclude that routine post-extubation CXRs can have a useful role.