On Three-Dimensional Reconstruction and Identification of Craniofacial Images---Development of a Software Platform

  • Wan, Shu-Yen (PI)
  • Huang, Chiung-Shing (CoPI)
  • Yu, Chung Chih (CoPI)

Project: National Science and Technology CouncilNational Science and Technology Council Academic Grants

Project Details

Abstract

Cephalometry has been well accepted as a radiographic technique for portraying the human head in terms of measurable geometry. It can be employed to depict the craniofacial morphology of a side, e.g., frontal or lateral, and to study the bony structure of the face for diagnosis, treatment planning/evaluation, and growth prediction and monitoring. The inherent limitation of cephalometry is, however, its two-dimensional nature which exhibits great variability in relationship measurements of cephalometric landmarks, such as the distance between gnathion and zygomaticotemporal points. Three-dimensional craniofacial imaging presents more promising and reliable measurements, and thus three-dimensional craniofacial reconstruction (3D-CFR). It may take multiple cephalometric images of different sides, determine the features that match and then perform alignment, and construct the corresponding three-dimensional surface of the face. With help of computed tomography (CT) and magnetic resonance imaging (MRI) technology, both the surface and the underlying hard tissue can be captured and visualized in three dimension. In this project, we propose computerized methods to perform image segmentation, feature extraction, landmark matching and identification, three-dimensional reconstruction from two-dimensional images of cephalometry or surface scan, and visualization. Several image segmentation techniques, such as symmetric region grow, mean shift, level set modeling, etc., will be examined and implemented. An improved robust and fast iterative closest point (ICP) algorithm will be proposed to hierarchically match and identify rigid and non-rigid cephalometric landmarks. We shall build a software platform to accommodate the implemented methods as components. Components are grouped together to perform certain task and is archived in a configuration script that records the sequence of the process and the corresponding parameter settings. The project will be carried out in three years. In the first year, we shall build the infrastructure and prototype of the software platform that enables the users to perform interactive manipulation and matching of two-dimensional cephalometric landmarks. Fundamental object-oriented image processing and analysis components are available in the first year. Late in the first year and in most of the second year, we shall realize the algorithms of symmetric region grow, level set modeling, mean shift, and semi-automatic extraction of features. All of these methods also serve as components that can be easily plugged into the proposed software platform. At the same time, development of the same components on Amira proceeds in parallel. Such additional implementation will facilitate the comparison of the proposed platform with the commercial Amira. Late in the second year and in most of the third year, we shall realize the improved ICP, perform matching and identification between images over time, or images from the normal and abnormal. Later in the third year, we shall perform unit/integration/system testing on the components and software platform with realistic image data in hope to disclose random errors and make system refinement. The scalability of the (component-based) software platform is also one of our major emphases to assure usability for further research on more biological and medical/clinical applications.

Project IDs

Project ID:PB9801-2178
External Project ID:NSC97-2221-E182-037-MY3
StatusFinished
Effective start/end date01/08/0931/07/10

Fingerprint

Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.