Biomedical Engineering Reference
In-Depth Information
which a user can modify the result of a segmentation in any way desired. There
are no limitations in shape, size, or complexity of a region under consideration.
The only requirement is that the given region have a spherical topology.
7.4
Conclusions
Image segmentation is an important component of any image analysis system.
In medical imaging, it is essential that an image is accurately segmented so
that different measurements about the region are accurately determined. In this
paper, the idea of using a computer-aided design system to effectively revise
the result of an automatically determined segmentation was introduced. In the
proposed system, a RaG surface is fitted to voxels representing a 3D region by
the least-squares method. The surface and the original volumetric image are then
overlaid and the surface is interactively revised until the desired segmentation
is achieved.
The system provides the option of using the output of an automatically ob-
tained segmentation as the input or manually creating an initial segmentation
by selecting a number of 3D points in the given image volume. In the latter case,
an initial surface is created from the points and overlaid with the image. The
user can then observe the image data and revise the surface to a desired shape.
Because a region of interest is represented by a parametric surface, the surface
may be sent to a computer-aided manufacturing system for construction of an
actual 3D model of the region.
7.5
Acknowledgements
The image used in Fig. 7.10 is from Georg Gotschuli and Erich Sorantin, Uni-
versity of Garz, Austria; the cardiac MR image in Fig. 7.9 is from David Turner,
Presbyterian-St. Luke's Hospital, Chicago, IL; and the MR brain image containing
only the brain in Fig. 7.9 is from Terry Oaks, University of Wisconsin, Milwaukee.
We appreciate all these contributions.
