Accuracy of automatic airway morphometry in computed tomography-correlation of radiological-pathological findings

Airway morphology shows characteristic changes in different pathologies. This study assesses the accuracy of a current automatic airway assessment technique by correlating CT images of porcine airways to histological slices of the same localization.

Four isolated and ventilated porcine lungs were frozen in a liquid nitrogen bath and examined with a CT scanner (MDCT). This technique both preserved normal radiomorphological appearance and made it possible to slice the specimens for histological examination for subsequent correlation. The parameters wall thickness (WT), wall percentage (WP), and total diameter (TD) were assessed by computer-aided measurement of the MDCT images using an integral-based method (IBM) and by manually measuring the histological slices with an electronic caliper.

The radiological-pathological correlation could be performed for 16 localizations. Mean relative errors for WT, WP, and TD were 11%, 5.6%, and 8.5%, respectively. Correlation was very high with coefficients r of 0.951 for WT, 0.916 for WP, and 0.987 for TD.

Our results are comparable to previously described errors in phantom correlations but are the first proof of ex vivo feasibility. Thus, by applying this freezing technique to MDCT data of diseased, explanted lungs and by combination with the IBM, further experiments can be performed to explore the effects of airway pathology on radiological morphology.