Information Technology Reference
In-Depth Information
oligonucleotides [1] [4]. The data obtained from microarrays are an important source
of knowledge to prevent and detect cancer. The analysis of this information allows the
detection of patterns that characterize certain diseases and, most importantly, the genes
associated with these different diseases. Since the amount of data obtained from mi-
croarrays is huge and the time required to analyze the data is very high, it is necessary
to obtained novel computational techniques that can provide automatic processing and
artificial intelligence techniques that provide behaviours similar to the human ones.
An expression analysis basically consists of three stages: normalization and filter-
ing; clustering and classification; and extraction of knowledge. These stages are car-
ried out from the luminescence values found in the probes. Presently, the number of
probes containing expression arrays has increased considerably to the extent that it
has become necessary to use new methods and techniques to analyze the information
more efficiently [5]. It is necessary to develop new techniques to analyze large vol-
umes of data, extract the relevant information, and delete the information which has
no relevance to the classification process. Moreover, the knowledge obtained during
the classification process is of great importance for subsequent classifications. There
are various artificial intelligence techniques such as artificial neural networks [6] [7],
bayesian networks [8], and fuzzy logic [9] which have been applied to microarray
analysis. While these techniques can be applied at various stages of expression analy-
sis, the knowledge obtained cannot be incorporated into successive tests and included
in subsequent analyses.
The system proposed in the context of this work focuses on the detection of car-
cinogenic patterns in the data from microarrays for patients, and is constructed from a
CBR system that provides a classification technique based on previous experiences
[11]. The system is an evolution of our previous work in the classification of leukemia
patients [12], where a mixture of experts was used. The incorporation of the CBR
paradigm to health care [13] [14] provides additional learning and adaptation capabili-
ties. Moreover. The filtering and extraction of knowledge models have been improved
and new techniques have been incorporated. The purpose of case-based reasoning
(CBR) is to solve new problems by adapting solutions that have been used to solve
similar problems in the past [10]. A CBR manages cases (past experiences) to solve
new problems. The way cases are managed is known as the CBR cycle, and consists
of four sequential steps which are recalled every time that a problem needs to be
solved: retrieve, reuse, revise and retain. Each of the steps of the CBR life cycle re-
quires a model or method in order to perform its mission.
The approach presented in this work focuses on the classification of subtypes of
leukemia, specifically, to detect patterns and extract subgroups within the CLL type
of leukemia, and incorporates various techniques of computational intelligence at
different stages of the reasoning cycle of a CBR system. In the retrieve phase, new
pre-processing and filtering techniques are incorporated in order to select the probes
with relevant information for classifying patients. This innovative method notably
reduces the dimensionality of the data, which makes it possible to use techniques with
greater computational complexity in later stages of the CBR cycle, which would oth-
erwise be unviable. The reuse stage incorporates a classification technique based on
ESOINN [15] neural networks, that proposes a novel method for generating clusters,
and for identifying the nearest cluster for the final classification. An additional group-
ing technique known as PAM [16] (Partition around medoids), is also used, resulting
Search WWH ::




Custom Search