Biomedical Engineering Reference
In-Depth Information
Figure 4.
Neighborhood search for the minimum energy configuration.
process of segmentation. In particular, ideally, the user has to provide an initial
contour near the desired edge. The snake deforms under the action of the local
image forces and geometric constraints until it conforms to the final edges of the
image.
The deformable model is in itself not free from limitations. In the original
proposition, the user needs to provide the initial contour very near the desired edges;
otherwise, the snake will not be able to deform to capture the desired anatomi-
cal structure. At the initial stage a number of solutions [17, 18] were provided in
different forms to allow the snake more evolution. Cohen et al. [13] proposed
a solution to propagate the contour faster toward the desired image edges. An
internal pressure force was introduced by regarding the curve or surface as an in-
flating
balloon
. This pressure pushes the contour boundary toward the edges, and
thus makes initialization of the snake a simpler process. However, the associated
limitation of the snake remains in its ability to balance the strength of the balloon
force with edge strength. As the balloon force is increased, there is a chance of
leakage at weak edges. The addition of balloon pressure, though, adds to the prop-
agation strength; however, this increases the instability of the snake framework.
Berger et al. [18] proposed a “snake-growing” algorithm, where the snake grows
based on the local contour information. Figure 6 shows the comparative result of
a snake-growing compared to a conventional snake.
