International Journal of Applied Engineering Research ISSN 0973-4562 Volume 11, Number 6 (2016) pp 4223-4229
© Research India Publications. http://www.ripublication.com
4223
An enhanced k nearest neighbor method to detecting and classifying MRI
lung cancer images for large amount data
P. Thamilselvan
Research Scholar, Department of Computer Science, Bishop Heber College (Autonomous),
Tiruchirappalli, Tamilnadu, India.
Dr. J. G. R. Sathiaseelan
Head, Department of Computer Science, Bishop Heber College (Autonomous),
Tiruchirappalli, Tamilnadu, India.
Abstract
The k nearest neighbor classification method is one of the
humblest method in conceptually and it is a top method in
image mining. In this work, the enhanced k nearest neighbor
(EKNN) technique has been implemented to identify the
cancer and automatic classification of benign and malignant
tissues in the huge amount of lung cancer image datasets. In
this proposed system, we have used three stages such as
preprocessing, identification of cancer and classification. In
preprocessing phase, the morphological method has been used
to improve the quality of images. We used enhanced k nearest
neighbor (EKNN) classifier for identifying the cancer and
classifying the images. The classification is done by
implementing four steps of k nearest neighbor which are
calculated based on Euclidean distance, define the k value,
assigning majority class and finding the minimum distance.
The nodule identification and classification of process trained
and tested on large-scale image databases.
Keywords: Image mining, Image Classification,
Morphological Method, K Nearest Neighbor, MRI lung
images, large amount of data, Classification Accuracy
Introduction
Image mining is the calculation progression of removing
implicit nontrivial unknown patterns in large amount of data
repository. The traditional image mining algorithms are being
modified and applied in different research field such as
finance, computer security, web content mining, medical and
fault diagnosis. Classification is a one of the predominant task
in image mining, it also learns to classify the pattern with the
help of training process data. The lung cancer is a system of
cancer that has become a substantial reason of worldwide
death based on the existing reports [1]. The CAD (Computer
Aided Diagnosing) system applications that contain
mammography masses [2], coronary artery disease [3],
various types of cancer like lung, breast, colon [4-6]. The
cause of cancer that rests early identification of cancer is the
greatest promising ways to decrease the number of deaths. In
order to diagnose medical image modalities such as MRI
images, CT (Computed Tomography) images and
mammography have been implemented to identify the
abnormality images. In this work, we have enhanced k nearest
neighbor method for detecting and classifying the magnetic
resonance lung cancer images.
The k nearest neighbor classification rule is based on the
density evaluation based on the distance of nearest neighbors
and it is a non-parametric technique which is used for
classification. The k nearest neighbor classifier method is
implemented to classify the handler’s activity based on the
structures. In nearest neighbor classification method, the
number of modules and structure selection will be key factors.
This method for classifying test examples based on nearest
training examples in feature space [7]. Basically Euclidean
distance vector is used to calculate the familiarity of the
samples. In this work, we concentrate on image classification
based on the nearest neighbor method to classify medical
images. Image classification is prevalent research area in the
field of computer visualization [8-11]. Basically an image
classification process can be divided into two category i. e.
non parametric and parametric classes. The medical images
play a vital role to identify the various types of syndrome in
human beings. Particularly, magnetic resonance images will
be very useful to identify the formation of cancers in human
lung image, brain image and breast image treatments. The
evaluation of those images with proper classification methods,
it will tell valuable information to neurosurgeons whether the
level of the formation of cancer in the human lung is benign
and malignant [12]. This research mainly concentrated
detecting lung cancer tissues in large amount of MRI image
data set and measuring the performance proposed EKNN
method based on the distance of nearest neighbor and
minimize the processing time.
Related Work
Sreeparna et al. [13] suggested decision tree method for
detecting and classifying the retinal abnormalities. The
decision tree method is implemented to hybrid contextual
evidence with images acquired from other database images.
The abnormalities of the retinal images which have been
considered in this study from abnormalities such as micro
aneurysm, diabetic retinopathy, age related macular
degeneration, arising from diabetic retinopathy, hard exudates
and cotton wool spots. The decision tree method is a decision
support algorithm that used to divide a collection of cases into
homogeneous groups and it has like structure or flow chart.