Bio-medical Ontologies Maintenance and Change Management

Multimedia Medical Databases 123
A simple classifier is the nearest-neighbor classifier, where each pixel or voxel
is classified in the same class as the training data with the closest intensity. The knearest-neighbor
(kNN) classifier is a generalization of this approach, where the
pixel is classified according to the majority vote of the k-closest training data [76].
Classifier methods have some advantages: are relatively computationally efficient
and can be applied to multi-channel images.
The disadvantages are:
• They generally do not perform any spatial modeling
• The requirement of manual interaction for obtaining training data; training sets
can be acquired for each image that requires segmenting, but this can be time
consuming and laborious.
• Using of the same training set for a large number of scans can lead to results
which do not take into account anatomical and physiological variability between
different subjects.
Clustering algorithms work as classifier methods but they don’t use training
data. As a result they are called unsupervised methods. Because there isn’t any
training data, clustering methods iterate between segmenting the image and characterizing
the properties of the each class. It can be said that clustering methods
train themselves using the available data [76]. The next three commonly used
clustering algorithms must be mentioned:
• K-means or ISODATA algorithm (clusters data by iteratively computing a
mean intensity for each class and segmenting the image by classifying each
pixel in the class with the closest mean)
• The fuzzy c-means algorithm (it is a generalization of the K-means algorithm;
it is based on fuzzy set theory)
• The expectation-maximization (EM) algorithm (works with the same clustering
principles and is based on the assumption that the data follows a Gaussian mixture
model) [3, 10]
These algorithms have the following disadvantages: they require an initial
segmentation and do not directly incorporate spatial modeling and can therefore
be sensitive to noise and intensity in homogeneities.
Markov random field (MRF) is a statistical model that can be used within segmentation
methods. For example, MRFs are often incorporated into clustering
segmentation algorithms such as the K - means algorithm under a Bayesian prior
model. MRFs model spatial interactions between neighboring or nearby pixels. In
medical imaging, they are typically used to take into account the fact that most
pixels belong to the same class as their neighboring pixels. In physical terms, this
implies that any anatomical structure that consists of only one pixel has a very low
probability of occurring under a MRF assumption [76].
The disadvantages are:
• Proper selection of the parameters controlling the strength of spatial interactions
is necessary
• Usually require computationally intensive algorithms