Extrapolation in the analysis of lung aeration by computed tomography: a validation study.

TitleExtrapolation in the analysis of lung aeration by computed tomography: a validation study.
Publication TypeJournal Article
Year of Publication2011
AuthorsReske, Alexander, Rau Anna, Reske Alexander, Koziol Manja, Gottwald Beate, Alef Michaele, Ionita Jean-Claude, Spieth Peter, Hepp Pierre, Seiwerts Matthias, Beda Alessandro, Born Silvia, Gerik Scheuermann, Amato Marcelo B. P., and Wrigge H.
JournalCrit Care
AbstractINTRODUCTION: Computed tomography (CT) is considered the gold standard for quantification of global or regional lung aeration and lung mass. Quantitative CT, however, involves the exposure to ionizing radiation and requires manual image processing. We recently evaluated an extrapolation method which calculates quantitative CT parameters characterizing the entire lung from only ten reference CT-slices thereby reducing radiation exposure and analysis time. We hypothesized that this extrapolation method could be further validated using CT-data from pigs and sheep, which have a different thoracic anatomy. METHODS: We quantified volume and mass of the total lung and differently aerated lung compartments in 168 ovine and 55 porcine whole-lung CTs covering lung conditions from normal to gross deaeration. Extrapolated volume and mass parameters were compared to the respective values obtained by whole-lung analysis. We also tested the accuracy of extrapolation for all possible numbers of CT slices between 15 and 5. Bias and limits of agreement (LOA) were analyzed by the Bland-Altman method. RESULTS: For extrapolation from ten reference slices, bias (LOA) for the total lung volume and mass of sheep were 18.4 (-57.2 to 94.0) ml and 4.2 (-21.8 to 30.2) grams, respectively. The corresponding bias (LOA) values for pigs were 5.1 (-55.2 to 65.3) ml and 1.6 (-32.9 to 36.2) grams, respectively. All bias values for differently aerated lung compartments were below 1% of the total lung volume or mass and the LOA never exceeded +/-2.5%. Bias values diverged from zero and the LOA became considerably wider when less than ten reference slices were used. CONCLUSIONS: The extrapolation method appears robust against variations in thoracic anatomy which further supports its accuracy and potential usefulness for clinical and experimental application of quantitative CT.