Information Technology Reference
In-Depth Information
Besides, FCM has been combined with other segmentation techniques as dy-
namic contours [2] for the segmentation of thalamus from brain in MRI. FCM
has been also mixed with the FHCE frequency histogram of connected elements in
mamogrphic microcalcification detection [7] and combined with Markov Random
Fields for the detection of brain activation regions in Functional Magnetic Reso-
nance Imaging (fMRI) [17].
The fact that in the standard fuzzy c-means no spatial context information is taken
into account makes it sensitive to noise. Therefore there are a lot of proposals that
try to solve this problem such as [30]. Other works modify the effective objective
function of fuzzy c-means by replacing the Euclidean distance with kernel-induced
distance [21] or with a robust kernel-induced distance [19] for clustering a corrupted
dataset of breast and brain medical images. In [38] a spatial penalty on the mem-
bership functions and a kernel-induced distance metric for MRI segmentation are
proposed.
12.4.2
Fuzzy Connectedness
The notion of fuzzy connectedness assigns a strength of connectedness to every
possible path between every possible pair of image elements. This concept leads to
powerful image segmentation algorithms [41], [34].
This methodolgy has been extensively used in medical image segmentation, for
example [39] present a method to extract abdominal organs using a presegmented
atlas. Another application was presented in [27] to segment phantom images, MRI,
computed tomography, and infrared data. Fuzzy connectedness was combined with
and edge detection for knee tissues in CT image and segmentation of brain tissues
in MRI image [24].
Fuzzy conectedness has also been extended to combine methods with the concept
of membership connectedness [16].
12.4.3
Fuzzy Rule Based Systems
Fuzzy rule based systems have been used to model the relations between body or-
gans or other regions of interest by rules.
Identifying multiple abdominal organs or brain regions from medical image se-
ries is necessary for the diagnosis. To address the issue of high variations in or-
gan position, shape and consecutive organ region overlap constraints, spatial fuzzy
rules and fuzzy descriptors have been adopted in several papers [23]. Sometimes,
the necessity of detecting a set of organs is called model based structural pattern
recognition. In these techniques, models and data structures are based on the use of
fuzzy restrictions. In [31] is presented a segmentation method of the coronary artery
tree in x-ray angiographic images, based on a fuzzy structure pattern inferring. In
[14] the role of the fuzzy logic is to fuse the voxel intensity information from the
time of flight angiography with the corresponding vesselness information based on a
designed rule base.
