The upper airways are a major site of congenital and acquired pediatric airway obstruction. Airway size information can
be used for pre-surgical planning and post-treatment assessment. The aim of this research is to develop a greater
understanding of the growth and variation of the normal pediatric airway to assist a surgeon in optimizing the outcomes for patients by increasing efficiency and accuracy. The standard imaging tool for measuring airway geometry has been CT (computed tomography), but to eliminate the risks of radiation exposure during prospective studies, we have developed an image analysis system to measure airway geometry in MR (magnetic resonance) images of the upper
airway. Six adult patients that had CT and MR images of a normal airway were used as the training set to optimize the
segmentation cost function to find the appropriate 3D surface. 25 normal pediatric subjects were segmented and
measured and then compared to the segmentations of three experts. Dice similarity coefficient and boundary point
distances were used for comparison metrics. The automated segmentations correlated well and were not significantly
different from those from the group of experts. A group of 90 children were measured and plotted against age, gender, z-scores for weight and height. In an attempt to model growth, cross-sectional area showed a significant correlation with a 4th degree polynomial of age. This work demonstrates that MR imaging can be used for measuring the pediatric upper airway and to develop growth models to assist in pre-surgical planning.