Biomedical Engineering Reference
In-Depth Information
Chapter 7
A Computer-Aided Design System
for Segmentation of Volumetric Images
Marcel Jackowski
1
and Ardeshir Goshtasby
2
7.1
Introduction
Image segmentation is the process of partitioning an image into meaningful re-
gions. For the regions to be meaningful, they should represent objects or their
parts. Difficulties arise when properties within objects vary or boundaries of
objects become blurred. The problem is worsened when sensor inaccuracies
exist and noise is present in the image. These variations, which are often unpre-
dictable, make it impossible to develop an automatic method that can segment
all images correctly. Because a high accuracy is demanded in the segmenta-
tion of medical images, the user has a critical role in examining the results and
correcting the possible errors.
Image segmentation is perhaps the most studied area in image analysis. A
large number of papers on this topic is published annually in image analysis
journals and conference proceedings. The developed methods often take into
consideration various properties of images or objects, and when such properties
deviate from the anticipated ones, errors occur. Even for a limited class of
images, for instance MR brain images, various methods have been developed,
none of which is guaranteed to work correctly on a new image. This may be
because there are sensor variations; variations in the brain's shape, size, and
