Three-dimensional reconstruction of soft palate modeling from subject-specific magnetic resonance imaging data

The objectives of this study were to extract a computational three-dimensional (3D) soft palate model from a set of magnetic resonance imaging (MRI) data and to identify an approach that generates a patient-specific model in a computerized visual platform.

Multiple MRI slices of the head and neck region of a young, non-overweight Caucasian male volunteer were taken in the supine position with a passive oral appliance in place. The DICOM (Digital Imaging and Communications) MRI slices were registered into a high-resolution volumetric data set for manually segmentation to generate a surface mesh and, with additional editing, a volume mesh. For biomechanical dynamic simulation and for physical simulation of the anatomical structures, the volume mesh format and multiple landmarks of each muscle were imported into ArtiSynth, a 3D biomechanical modeling toolkit.

The segmented soft palate complex consisted of five groups of muscles: levator veli palatini, tensor veli palatini, palatoglossus, palatopharyngeous and musculus uvulae. The palatine tonsil between the pharyngopalatine and glossopalatine arches was included in the segmentation.

The same procedure was used to build up a generic reference model of the dentition, tongue, mandible and airway from a mixture of medical records (CT and dental casts) of the same subject. This manual segmentation method eliminated the common errors that occur from an automatic segmentation although it was more time-consuming. It remains a fundamental process for analyzing the dynamic interaction between anatomical components in the oral, pharyngeal, and laryngeal areas.