25-08-2017, 09:32 PM
Abstract—In this paper we propose a new CAD ( Computer
Aided Diagnosis ) system to classify patients with Alzheimer
disease. Five textural features proposed by Harlick are
extracted from the MRI scans which characterize the disease.
An enhanced CBA algorithm is used to classify the images as
Normal or Abnormal based on the rule set generated during
the training phase. The experiments were conducted on OASIS
dataset and result show that the proposed system classifies
with 98% accuracy, making association rules efficient for the
classification task.
Keywords- MRI, Alzhemier, Classification, Association
Rules, Mining
I. INTRODUCTION
AD is a neurodegenerative disease that ravages its
victims’ memory to affect their learning, cognitive, and
communicative abilities . Alzheimer’s disease (AD) affects
the central nervous system and begins in middle to late life,
and results in severe dementia and ultimately death. It is the
fourth major cause of death after heart diseases, cancer and
stroke. As awareness grows, more and more old people
earlier assumed to be afflicted by other diseases are now
being diagnosed as suffering from AD. The demographic
data show that presently there are about 18 million people
with dementia in the world. Approximately 70% of this
number is contributed by the developing countries. The
projected estimate for Europe alone is about eight million
and for United States about four million AD patients.
Epidemiological information from India is rather scanty.
Satishchandra et al. [1] reported the first case of familial AD
(FAD) in India. Another study carried out in Kerala found 66
cases of dementia among 2067 persons over the age of 60
years, a prevalence rate of 3.2%. Out of this, 58% had
vascular type while 41% had Alzheimer’s type dementia.
Surveys conducted in Tamil Nadu and other parts of India
also suggest that Indians are as prone to dementia as other
ethnic groups. Considering India’s 70 million elderly
population and increasing life expectancy above 60, the
number of AD cases is expected to go up and demands for
the health care of demented people will rise remarkably in
the new millennium.
MRI scans have become indispensable for assessing
many brain related disorders. The task of analyzing a large
amount MRI scans daily demanded by medical centers is
tiresome and radiologists should have some automatic tools
to support it. Thus a CAD should support the physician’s
work, improving the diagnosis task. MRI is generally
regarded as a superior tool for brain imaging, as compared
with CT, due to the absence of ionizing radiation, increased
imaging flexibility, and better tissue contrast. Unfortunately,
expense, patient claustrophobia, or the presence of metal
implants or medical devices common in older individuals
can limit the use of MRI.
Classification and Association rule mining has been
under study for quiet a long period. Building effective
classification systems is one of the central tasks of data
mining and machine learning. Past research has produced
many techniques (e.g. decision trees, rule learning and
Naïve-Bayse classification) and systems (e.g., C4.5 [2], CN2
[3], and RIPPER [4]). The existing techniques are, however,
largely based on heuristic/greedy search. They aim to find
only a subset of the regularities (e.g., a decision tree or a set
of rules) that exist in data to form a classifier. However the
potential of these classification techniques has not been fully
uncovered. In this paper we propose a two phase algorithm
with association rules enhanced classifier for classification of
MRI Scans.
Aided Diagnosis ) system to classify patients with Alzheimer
disease. Five textural features proposed by Harlick are
extracted from the MRI scans which characterize the disease.
An enhanced CBA algorithm is used to classify the images as
Normal or Abnormal based on the rule set generated during
the training phase. The experiments were conducted on OASIS
dataset and result show that the proposed system classifies
with 98% accuracy, making association rules efficient for the
classification task.
Keywords- MRI, Alzhemier, Classification, Association
Rules, Mining
I. INTRODUCTION
AD is a neurodegenerative disease that ravages its
victims’ memory to affect their learning, cognitive, and
communicative abilities . Alzheimer’s disease (AD) affects
the central nervous system and begins in middle to late life,
and results in severe dementia and ultimately death. It is the
fourth major cause of death after heart diseases, cancer and
stroke. As awareness grows, more and more old people
earlier assumed to be afflicted by other diseases are now
being diagnosed as suffering from AD. The demographic
data show that presently there are about 18 million people
with dementia in the world. Approximately 70% of this
number is contributed by the developing countries. The
projected estimate for Europe alone is about eight million
and for United States about four million AD patients.
Epidemiological information from India is rather scanty.
Satishchandra et al. [1] reported the first case of familial AD
(FAD) in India. Another study carried out in Kerala found 66
cases of dementia among 2067 persons over the age of 60
years, a prevalence rate of 3.2%. Out of this, 58% had
vascular type while 41% had Alzheimer’s type dementia.
Surveys conducted in Tamil Nadu and other parts of India
also suggest that Indians are as prone to dementia as other
ethnic groups. Considering India’s 70 million elderly
population and increasing life expectancy above 60, the
number of AD cases is expected to go up and demands for
the health care of demented people will rise remarkably in
the new millennium.
MRI scans have become indispensable for assessing
many brain related disorders. The task of analyzing a large
amount MRI scans daily demanded by medical centers is
tiresome and radiologists should have some automatic tools
to support it. Thus a CAD should support the physician’s
work, improving the diagnosis task. MRI is generally
regarded as a superior tool for brain imaging, as compared
with CT, due to the absence of ionizing radiation, increased
imaging flexibility, and better tissue contrast. Unfortunately,
expense, patient claustrophobia, or the presence of metal
implants or medical devices common in older individuals
can limit the use of MRI.
Classification and Association rule mining has been
under study for quiet a long period. Building effective
classification systems is one of the central tasks of data
mining and machine learning. Past research has produced
many techniques (e.g. decision trees, rule learning and
Naïve-Bayse classification) and systems (e.g., C4.5 [2], CN2
[3], and RIPPER [4]). The existing techniques are, however,
largely based on heuristic/greedy search. They aim to find
only a subset of the regularities (e.g., a decision tree or a set
of rules) that exist in data to form a classifier. However the
potential of these classification techniques has not been fully
uncovered. In this paper we propose a two phase algorithm
with association rules enhanced classifier for classification of
MRI Scans.