Background: Our goal was to evaluate computed tomographic (CT) scans of the chest as a means of stratifying smoke inhalation injury (SII) severity.
Methods: Twenty anesthetized sheep underwent graded SII: group I, no smoke; group II, 5 smoke units; group III, 10 units; and group IV, 16 units. CT scans were obtained at 6, 12, and 24 hours after injury. Each quadrant of each slice was scored subjectively: 0 = normal, 1 = interstitial markings, 2 = ground-glass appearance, and 3 = consolidation. The sum of all scores was the radiologist's score (RADS) for that scan. Computerized analysis of three-dimensional reconstructed scans was also performed, based on Hounsfield unit ranges: hyperinflated, -1,000 to -900; normal, -899 to -500; poorly aerated, -499 to -100; and nonaerated, -99 to +100. The fraction of abnormal lung tissue (FALT) was computed from poorly aerated, nonaerated, and total volumes. Mean gray-scale density (DENS) was also computed.
Results: SII resulted in severity- and time-related changes in oxygenation (alveolar-arterial gradient), ventilation (respiratory rate-pressure product), DENS, FALT, and RADS. Ordinal logistic regression generated a predictive model for severity of injury (r2 = 0.623, p = 0.001), retaining RADS at 24 hours and rejecting the other variables.
Conclusion: At 24 hours, CT scanning enabled SII severity stratification; qualitative evaluation (RADS) outperformed current semiautomated methods (DENS, FALT).