Extracting moving boundaries from dynamic, multislice CT images for fluid simulation

Abstract

A method for extracting solid boundary motion from dynamic CT images of a human oropharyngeal swallow is presented. The data-set has high spatial resolution () but low temporal resolution (10 Hz). The low temporal resolution results in motion artefacts (i.e. blurring, ghosting) due to the fast moving solid structures. Automated segmentation and registration methods were not successful, so a tool for manual 3D deformable registration was created as a plug-in for the open-source modelling software, Blender. The tool gives users a high level of control to sculpt a mesh to match 3D volumetric image data. The resulting series of meshes, each corresponding to an instance in time, can be used in a fluid simulation using a technique such as smoothed-particle hydrodynamics.