Semi-Automatic Volumetric Segmentation of the Upper Airways in Patients with Pierre Robin Sequence

Abstract

Pierre Robin malformation is a rare craniofacial dysmorphism whose pathogenesis is multifactorial. Although there is some agreement in non-invasive treatment in less severe cases, the dispute is still open on cases with severe respiratory impairment. We present a semi-automatic novel diagnostic tool for calculating upper airway volume, in order to eventually address surgery in patients with Pierre Robin Sequence (PRS). Multidetector CT datasets of two patients and two controls were tested to assess the proposed method for ROI segmentation, upper airway volume computation and three-dimensional reconstructions. The experimental results show an irregular pattern and a severely reduced cross-sectional area (CSA) with a mean value of 8.3808 mm2 in patients with PRS and a mean CSA value of 33.7692 mm2 in controls (a ΔCSA of about −75%). Moreover, the similarity indexes and sensitivity/specificity values obtained showed a good segmentation performance. In particular, mean values of Jaccard and Dice similarity indexes were 91.69% and 94.07%, respectively, while the mean values of specificity and sensitivity were 96.69% and 98.03%, respectively. The proposed tool represents an easy way to perform a quantitative analysis of airway volume and useful 3D reconstructions.