Biomedical Engineering Reference
In-Depth Information
hand in terms of anatomy of interest (e.g., endocardial surface of my-
ocardium muscle), imaging modality (e.g., three-dimensional real-time ul-
trasound) and clinical application targeted (e.g., quantification of volume).
Only in this context can a segmentation method be really tuned, tested,
and validated for clinical application [107].
Questions
1.
Define the principle idea of the level set framework for segmentation of an
image given an initial contour C and a speed function V
>
0
. What are the
advantages of using an implicit formulation of the problem?
2.
What are the limitations associated with gradient-based speed terms? Why
is it especially problematic with medical images?
3.
What is a regularization term? What is its main functionality? Give some
examples.
4.
What is the entropy principle for implementation of a level set deformable
model with finite difference? In what case does it apply?
5.
Explain the concept of speed extension for image-based speed terms? Why
is it necessary? Propose a simple algorithm to implement it.
6.
Is the standard level set framework preserving the distance function? Why
is this an important concept for segmentation applications?
7.
Why is there a need for reinitialization of the distance function?
8.
Outline in a flowchart the structure of an iterative level set segmentation
algorithm using a gradient-based speed term. Use a convergence criteria
(without detailing it) to stop the iterations.
9.
Design a level set segmentation algorithm for extraction of the endocardial
and epicardial surfaces of the left ventricle from an MRI volume? What are
the properties of the data that can be used to define the speed function? Is
there a way to perform simultaneous segmentation of the two surfaces?
