A computer-aided design system for three-dimensional medical image segmentation.

详细信息   

• 作者：Jackowski ; Marcel Parolin.
• 学历：Doctor
• 年：2001
• 导师：Goshtasby, Ardeshir
• 毕业院校：Wright State University
• 专业：Computer Science.;Engineering, Biomedical.
• ISBN：0493942262
• CBH：3074236
• Country：USA
• 语种：English
• FileSize：4203904
• Pages：136

文摘

Traditionally, computer-aided design (CAD) systems have been used for engineering design and drafting purposes. In this work, the concept of CAD systems has been extended to modeling of digital shapes obtained by segmenting 3-D medical images.;Digital shapes are obtained by segmenting 3-D medical images such as Magnetic Resonance (MR), X-Ray and Computer Tomography (CT) or Positron Emission Tomography (PET). By representing such shapes with 3-D surfaces, we not only allow visualization of complex image structures but also assist in erroneous segmentation results obtained by automated methods.;It is well known that automatic segmentation methods do not always correctly segment images at all times. This is due to variations in imaging sensor parameters, presence of noise, and variations in tissue properties. Therefore, design of automatic methods that can correctly segment medical images is very difficult, if not impossible. A clinician often finds it necessary to refine the result of an automatically determined segmentation.;The goal of this work is to develop a CAD system with which users can interactively edit a segmentation result obtained by an automated method. By representing digital shapes with a parametric surface, rendering can be performed smoothly and in shaded form, and modification of the shape can be done quickly and effectively. A parametric surface model based on rational Gaussian basis functions is used to represent image regions obtained by image segmentation methods.;The surface will be obtained by morphing a sphere to a given 3-D region with a prescribed accuracy. This surface can then be deformed in real time, giving the user the impression that an elastic object is revised to correct for segmentation inaccuracies. The revised surface can then be used to estimate shape properties, such as diameter, volume, and normals.;This information can be further used in the diagnosis of pathologies, to track progress of therapy, and to estimate transformation parameters in image registration.