Biomedical Engineering Reference
In-Depth Information
and to emphasize a state-of-the-art family of the model using shape and regional
information.
Aside from medical imagery, the diversity of applications of level sets has
reached into several fields. These applications and their relevant works include
[1]: geometry (Angenent et al. [2], Chopp et al. [3], and Sethian et al. [4]);
grid generation (Sethian and Malladi [5]); fluid mechanics (Sethian et al. [6] and
Sussman et al. [7]); combustion (Rhee et al. [8]); solidification (Sethian and
Strain [9]); device fabrication (Adalsteinsson and Sethian [10]); morphing (Breen
and Whitaker [11]); object tracking/image sequence analysis in images (Mansouri
and Konrad [12], Paragios and Deriche [13] and Kornprobst et al. [14]); stereo
vision (Faugeras and Keriven from INRIA [15]); shape from shading (Kimmel et
al. [16, 17]); mathematical morphology (Sapiro et al. [18] and Sochen et al. [19]);
color image segmentation (Sapiro et al. [20]); 3D reconstruction and modeling
(Caselles et al. [21]); surfaces and level sets (Chopp [3] and Kimmel et al. [22]);
topological evaluations (DeCarlo and Gallier [23]); inpainting (Telea [24]); 2D
and 3D medical image segmentation (which will be discussed in more detail in the
following).
Segmentation has always been a critical component in two-dimensional (2D)
and three-dimensional (3D) medical imagery since it greatly assists in the process
of medical therapy [1]. The applications of shape recovery have been increasing
as scanning methods have become faster, more accurate, and less artifacted [25].
The recovery of the shapes of the human body is more difficult compared to other
imaging fields. This is primarily due to the great shape variability involved, the
complexity of medical structures, several kinds of artifacts, and restrictive body
scanning methods.
Since the introduction of the level set method, the concept of deformable
models for image segmentation defined in a level set framework has motivated
the development of geometric active contours. Much work on 2D and 3D seg-
mentation has been done around this method. For example, Malladi and Sethian
[26] proposed a real-time algorithm for medical shape recovery using the level
set method. Application of level sets for cortex unfolding was studied Faugeras
and coworkers from INRIA [27]. Application of the level set technique in cell
segmentation was introduced by Sarti et al. [28]. Niessen et al. [29] worked on
application of geodesic active contours for cardiac image analysis. There are also
many applications in brain imaging, like the work on gray matter/white matter
(GM/WM) boundary estimation by Gomes and Faugeras [30], GM/WM boundary
estimation with fuzzy models by Suri [31], and GM/WM thickness estimation by
Zeng et al. [32]. Angelini et al. [33] and Lin et al. [34] worked on 3D ultrasound
image analysis using a three-dimensional level set.
As applications are becoming more and more widely used in medical image
segmentation and relative analysis, the challenges facing researchers are becoming
more difficult and complex. Initially, a level set without complicated regularizers
was utilized to solve the simple case where medical images had good contrast,
