Biomedical Engineering Reference
In-Depth Information
(a)
(b)
(c)
Figure 15.
Segmentation using a region-based active contour: (a) original image; (b) image
with initial contour; (c) image with segmentation result. See attached CD for color version.
and convoluted ventricular shapes when integrating the third spatial dimension.
From this work the reader will also find out how to assign parameters for specific
situations.
5.
IMAGE SEGMENTATION USING SHAPE PRIOR
Missing or diffuse boundaries can be a very challenging problem in medical
image processing, which may be due to patient movement, a low signal-to-noise
(SNR) ratio of the acquisition apparatus, or blending with similar surrounding
tissues. Under such conditions, without an a prior model to constrain the segmen-
tation, most algorithms (including intensity- and curve-based techniques) fail—
mostly due to the underdetermined nature of the segmentation process. Similar
problems arise in other imaging applications as well, and they also hinder segmen-
tation of an image. These image segmentation problems demand the incorporation
of as much prior information as possible to help the segmentation algorithms ex-
tract the tissue of interest.
A number of model-based image segmentation algorithms in the literature
were used to deal with cases when boundaries in medical images are smeared or
missing. In 1995, Cootes et al. [55] developed a parametric point distribution
model (PDM) for describing the segmenting curve by using linear combinations
of the eigenvectors that reflect variations from the mean shape. The shape and pose
parameters of this point distribution model were determined to match the points to
strong image gradients. Wang and Staib [60] developed a statistical point model
in 1998 for the segmenting curve by applying principal component analysis (PCA)
to the covariance matrices that capture the statistical variations of the landmark
points. They formulated their edge-detection and correspondence-determination
problem in a maximum a posteriori Bayesian framework. The image gradient
was used within that framework to calculate the pose and shape parameters that
