3D Livewire Based Segmentation and Modeling of Soft Palate

Obstructive Sleep Apnea (OSA) is a common disorder among North Americans and a major topic in health care nowadays. Though the exact cause of this disease is unclear, it is understood as partial or complete obstruction of airway due to collapsing of soft palate while sleeping. Our work is aimed at creating a reference model for soft palate; medical specialists can use a dynamic version of this model to study the origin and possible treatments for OSA or other swallowing disorders. This article deals with the design and implementation of a complete workflow/scheme for segmentation and modeling of soft palate using MR (Magnetic Resonance) images. The segmentation process relies on a semi-automatic technique called 3D Livewire. This method performs user-guided segmentation in order to balance automation and human intervention. It requires the use to draw contours in any two orthogonal directions of 3D medical data and can automatically create seeds in the third direction. The contours or crosssection curves thus extracted are in turn used to reconstruct the surface of the soft palate, which produces a smooth 3D model. The proposed workflow has shown encouraging results. The basic soft palate model created can be improved further to study its behavior and can be used to reproduce sleep apnea phenomenon in future.