An edge-driven 3D region-growing approach for upper airway morphology and volume evaluation in patients with Pierre Robin sequence. (2015)

• Record Type: Journal Article
• Authors: Militello, Carmelo; Vitabile, Salvatore; Rundo, Leonardo; Gagliardo, Cesare; Salerno, Sergio
• Is Part Of: International journal of adaptive and innovative systems. Volume 2:Number 3(2015)
• Page Start: 232
• Page End: 253
• Publication Date: 2015
• Subjects: 3D region growing -- edge-driven segmentation -- airway segmentation -- Pierre Robin sequence -- PRS -- 3D modelling -- medical DSS -- decision support systems -- MDSS -- upper airway morphology -- volume evaluation -- breathing difficulties -- feeding difficulties -- medical imaging -- regions of interest -- ROIs -- image segmentation -- healthcare technology; Adaptive computing systems -- Periodicals; Computer systems -- Technological innovations -- Periodicals; Systems software -- Periodicals; 004
• Languages: English
• ISSNs: 1740-2107

Abstract

Pierre Robin sequence (PRS) is a pathological condition responsible for a sequence of clinical events, such as breathing and feeding difficulties, that must be addressed to give the patient at least a chance to survive. By using medical imaging techniques, in a non-intrusive way, the surgeon has the opportunity to obtain 3D views, reconstruction of the regions of interest (ROIs), useful to increase understanding of the PRS patient's condition. In this paper, a semi-automatic approach for segmentation of the upper airways is proposed. The implemented approach uses an edge-driven 3D region-growing algorithm to segment ROIs and 3D volume-rendering technique to reconstruct the 3D model of the upper airways. This method can be used to integrate information inside a medical decision support system, making it possible to enhance medical evaluation. The effectiveness of the proposed segmentation approach was evaluated using Jaccard (92.1733%) and Dice (94.6441%) similarity indices and specificity (96.8895%) and sensitivity (97.6682%) rates. The proposed method achieved an average computation time reduced by a 16x factor with respect to manual segmentation.