Medical Image segmentation using Image Mining concepts

Scientific Journal Impact Factor (SJIF): 1.711
International Journal of Modern Trends in Engineering
and Research
www.ijmter.com
@IJMTER-2014, All rights Reserved 113
e-ISSN: 2349-9745
p-ISSN: 2393-8161
Medical Image segmentation using Image Mining concepts
Anirudh Vyas1
, Sitesh Kumar Sinha2
, Mukesh Kumar3
1
Research Scholar Computer Science Department, AISECT University Bhopal, India
2
Dean, AISECT University Bhopal, India
3
Assistant Prof. Computer Science Department, AISECT University Bhopal, India
Abstract— Image differencing is usually done by subtracting the low-level skin texture like strength
in images that are already associated. This paper extracts high-level skin texture in order to find out
an efficient image differencing method for the analysis of Brain Tumor. On the other hand, this
produces sets of skin texture that are both spatial. We demonstrate a technique that avoids arbitrary
spatial constraints and is robust in the presence of sound, outliers, and imaging artifact, while
outperforming even profitable products in the analysis of Brain Tumor images. First, the landmark
are establish, and then the top entrant are sorted into a end set. Second, the top sets of the two
descriptions are then differenced through a cluster judgment. The symmetry of the human body is
utilized to increase the accuracy of the finding. We imitate this technique in an effort to understand
and ultimately capture the judgment of the radiologist. The image differencing with clustered
contrast process determines the being there of Brain Tumor. Using the most favorable features
extracted from normal and tumor regions of MRI by using arithmetical features, classifiers are used
to categorize and segment the tumor portion in irregular images. Both the difficult and preparation
phase gives the proportion of accuracy on each parameter in neural networks, which gives the idea to
decide the best one to be used in supplementary works. The results showed outperformance of
algorithm when compared with classification accuracy which works as shows potential tool for
classification and requires extension in brain tumor analysis.
Keywords— Images, MMR, MCR, Segmentation etc.
I. INTRODUCTION
In this paper we discuss how to solve a medical image classification task, where depicted objects
of the same kind have to be classified either normal or diseased, and there are only differences in
local texture, often ill-defined, between objects from both classes. Previous pattern recognition
approaches to this problem are based on pixel or region classification and subsequent fusion of local
posterior probabilities to obtain an overall decision for the image under consideration. These
approaches require pixel or region labels to train local classifiers. For example, to detect interstitial
disease on chest radiographs pixel- and region-based classification approaches were applied in [1]
and [2] respectively. In practice, a pixel or region ground truth is often unavailable, or is unreliable
for ill-defined lesions. A ground truth on the image level, on the other hand, is almost always
available during the collection of a data set, or is easier to obtain. Therefore we aim at a classification
approach that allows us to classify an image as a whole from only overall image labels. Since the
information that concerns the presence or absence of pathology is local, an image representation is
introduced where global image features are derived from local per-pixel features. The starting point
of the method is the extraction of local features from spatially corresponding pixels in all images
under consideration. One way to obtain corresponding pixels across images is to warp a mean image
to all images. We segment an image to get a number of fixed landmarks that can be used to establish

International Journal of Modern Trends in Engineering and Research (IJMTER)
Volume 01, Issue 05, [November - 2014] e-ISSN: 2349-9745, p-ISSN: 2393-8161
a warping function. Next, a new set of image features is derived from local features of that image and
local features of another image, which we call a reference image. Features are calculated as distances
between two vectors whose elements are values of a certain per-pixel feature in all corresponding
points from a given image and a reference image. In this feature space supervised classification of
images can be performed using only global class labels of training images. With several reference
images, different image representations can be constructed and a pool of classifiers can be trained.
Then, for an unseen image, it is possible to combine multiple classification opinions in order to
smooth over mistakes of individual classifiers. The experiments in this paper focus on a medical
classification task, but the framework can be used for any image classification task where only
overall image class labels are available. In other domains, such as object detection, this type of task
is also recognized to be important (e.g. in [3]) and often referred to as weakly labeled image data. In
the next Section, the method is described in detail. Section 3 presents experimental results. Finally,
Section 4 provides a discussion and conclusions.
II. METHOD
Further in this section we assume that the classification task relates to an area of interest within an
image, and the word ”shape” is used to denote such an area. Shapes of the same kind are aimed to be
classified, e.g. lung fields in chest radiographs. Image representation and pixel correspondence
For our method it is important to establish the correspondence between points within analogous
shapes on two different images. Theoretically, various approaches to that might be considered. We
describe here an approach that uses Active Shape Models (ASM) for segmentation (see [4]). After
segmenting each shape X using ASM, positions of a number of corresponding landmark points on
the outlines of images become available. From a large set of images, the mean position of these
points can be computed. An image with these mean points is called a mean image. Next, a warping
function is determined between an image I containing a shape X, and the mean image by demanding
that the mean points are warped to the ASM landmark points in the image I. We use a warping
algorithm that is described in [5]. A warping function finds for a point in the mean image a related
point in the image I. We consider the following representation of X. Assume, that a number of
features M are extracted per pixel i, i ∈ X. A pixel i is represented by a feature vector xi = (xi1, xi2. .
. xiM). Thus we get a matrix representation of X:
III. BRAIN TUMOR
For detecting brain tumor by Segmentation model, a similar method in feature extraction approach is
used in both. [41,42] Fourier descriptors are used for precise extraction of boundary feature of the
Object Base
Classification
Pixel Base
Classification
Appling
Maximum
Likelihood
Supervised
Classification
Computing
Attributes
Image
Segmentation
Merging
Segments
Collecting
Training Sites
Processing

tumor region. But, these features introduce a large number of feature vectors that may cause the
problem of over learning or misclassification. Accordingly, these works are followed by GA as a
feature selection step. GA develops an efficient searching algorithm to combine these features to few
significant ones that are capable of representing that particular class. In other kinds of features are
applied for brain tumor detection. Six textural features based on GLCM have been used. GLCM {P
(d, q) (i, j)} represents the probability of occurrence of a pair of gray-levels (i, j) separated by a given
distance d at an angle q. The features such as contrast, inverse difference moment, correlation,
variance, energy and entropy are used. In principle component analysis (PCA) is used as a feature
extraction algorithm of brain tumor classification by use of Segmentation. PCA is one of the most
successful techniques that have been used in image recognition and comparison. Image in this work
is brain magnetic resonance image. Figure illustrates 5 division of brain tumor that are used.
Five classes of brain tumor
IV. SEGMENTATION AND FEATURE EXTRACTION ALGORITHMS
Image segmentation plays a crucial role in many medical imaging applications, especially in medical
image categorization. Because of this, correct features are extracting from a segmented
representation. In addition, in some works after feature extraction, an extra step is necessary to select
appropriate features among a large feature set. According to work in [4], feature selection is
important for medical data mining to reduce processing time and to increase classification accuracy.
According to the experiments in which the Segmentation model is used in the classification part,
dissimilar diseases such as cancer, eye sickness, skin, brain tumor especially, breast cancer can be
diagnosed.
In addition, the severity of diseases can be determined by the use of the image classifier. In this
section, according to the type of diseases, the methods of segmentation and feature extraction are
described.
Image segmentation and its performance evaluation are vital aspects in image dispensation because
of the complexity of the medical images; segmentation of medical descriptions remains a challenging
predicament. The Segmentation model can be used as an approach for segmentation of different
diseases such as breast cancer renal calculi and, especially, brain tumor.
The layout of this paper is as follows: introduces the Segmentation structure and its learning
algorithm.
V. ECCENTRICITY CALCULATION EQUATION
Eccentricity (E) has been widely used as a shape feature. It illustrates the way in which the region
points are scattered around the Centre of the image region. E is defined as the ratio of the major axis
of the region to the minor axis. It is calculated using central moments.

Segmentation
Image segmentation and its performance evaluation are vital aspects in image dispensation and, as
the difficulty of the medical images; segmentation of medical imagery remains a challenging
problem. The Segmentation model can be used as an approach for segmentation of different diseases
such as breast cancer renal calculi and, especially, brain tumor.
The layout of this paper is as follows: introduces the Segmentation structure and its learning
algorithm.
In segmentation and feature extraction algorithms section, the technique of segmentation and feature
extraction in medical image applications in which the segmentation copy is applied as classifier are
reviewed. Image classifier sector, presents a survey on works that used the segmentation model as
classifier in medical image classification and a brief comparison with other classifiers is proposed. In
order to clear the aim of this work, the outline of medical analysis by using the segmentation model
is presented. The main purpose of this work can be followed by the solid line used in segmentation.
The approach is used in image segmentation. The first case is when ground truth is available and the
other is when not available.
• When ground truth is available, segmentation can be evaluated in terms of accuracy and
robustness. The accuracy reacts the precision of segmentation with respect to ground truth. The
robustness is related to the accuracy degradation with respect to the degradation of the quality of
test data. There are various evaluation schemes.
a) Parameter based evaluation
b) Boundary based evaluation
c) Region based evaluation
• When ground truth is not available, evaluating segmentation results is similar to evaluating
cluster validity. Evaluate segmentation in terms of the contrast between the heterogeneity among
different regions versus the homogeneity within individual regions. Exact quantitative measure is
difficult in this case.
Parameter used for pixel
Mismatching Rate -It denotes how many percentage of labels differ in two labeled matrices
obtained by k Means classification technique
• Misclassification Rate (MCR) –The number of misclassified pixel divides by total number of
pixels multiply with 100
• MCR = (Number of misclassified pixel/ Total number of pixels)*100
• Computational Time: Total time CPU taken by proposed method for computation

VI. RESULTS
6.1 Misclassification Rate between Window Sizes:
Image_name MCR_2×2 MCR_3×3
Image 1 0.70 0.72
Image2 1.89 1.90
TABLE 1 MCR between ground truth & segmented Image
GRAPH1 Misclassification rate between Window Size
6.2 Mismatching Rate
Table 2 contains the mismatching rate between segmented image and ground truth image.
Image_name MMR_2×2 MMR_3×3
Image 1 1.12 1.56
Image2 0.56 1.23
TABLE 2 MMR between ground truth & segmented Image
6.3 Mismatching Rate between Shape

Graph 2 Represent the Mismatching Rate between Shape:
6.4 Time Comparison
Image_name
CPU_time_2×2
in sec
CPU_time_3×3
in sec
Image 1 77.767 82.74
Table 3 contains the CPU time:
Graph 3 Represent Time Comparison
6.5 APPLY NOISE:-
Image noise is the random variation of brightness or color information in images produced by
sensors and circuitry of a scanner or digital camera [8] .The principal sources of noise in digital
images arise during image acquisition (digitization) and/or transmission. There are different types of
noise. The blast used here are:
Gaussian Noise arises in an image due to factors such as electronic circuit noise and sensor noise due
to poor illumination.
– Salt and Pepper Noise is originate in location
6.5.1 Misclassification Rate between Window Sizes for Noisy image
Table 1 contain the misclassification rate between segmented image and ground truth image.

Image_name MCR_2×2 MCR_3×3
Image 1 2.16 2.99
Image2 1.11 1.67
6.5.2 CPU Time utilization after noice:
Table 3 contains the CPU time between segmented noisy image and ground truth image.
Image_name CPU_time_2×2
in sec
CPU_time_3×3
in sec
Image 1 121.35 95.65
Image2 79.23 95.326
Graph Represent the CPU time Utilization
VII. CONCLUSION & FUTURE WORK
In this work, we have tried to review the experiments in which Segmentation is applied as classifier
in the field of medical image categorization. In addition, special methods are used for segmentation
and feature mining steps in which applications that neural network is applied as classifier are
described. Although this work emphasizes on neural network as classifier, in adding up, it can be
used for other mission like denoising or segmentation. Neural network could be a good classifier for
medical image classification and assist physicians for earlier diagnosis of dissimilar diseases. These
classifiers conquer the disadvantage of fuzzy organization and neural networks by combining these
two smart methods. On the other hand, like every process, it has some drawbacks. Large effort data
could reduce the accuracy of its show. In addition, Neural Network performance decreases while
input samples are not enough and, also, there are many nodes in Neural. In general, the performance
of classifier depends on several factors such as size and quality of training position, the inflexibility
of the training introduce and also parameters selected to feed to Neural net as inputs can be authority
the classifier recital.
REFERENCES
[1] N. Senthilkumaran, "Genetic Algorithm Approach to Edge Detection for Dental X-ray Image Segmentation",
International Journal of Advanced Research in computer Science and Electronics Engeneering, vol.1, no.7,
2012.
[2] S. Binitha, S Siva Sathya, "A Survey of Bio inspired Optimization Algorithms", International Journal of Soft
Computing and Engineering (IJSCE), ISSN: 2231-2307, vol.2, Issue 2, May 2012.

[3] L. Florea, C. Florea, C. Vertan and A. Sultana, "Automatic Tools for Diagnosis Support of Total Hip
Replacement Follow-up ", Advances in Electrical and Computer Engineering, vol.11, no.4, pp.55-63,2011.
[4] T. Sag, M. Cunkas, "Development of Image Segmantation Techniques Using Swarm Intelligence", Proc. of the
1st Taibah University International Conference on Computing and Information Technology, pp.95-100, 2011.
[5] A. V. Alvarenga, "Artificial Ant Colony: Features and applications on medical image segmentation", Pan
American Health Care Exchanges Conference, pp. 96-101, 2011.
[6] J. Rahebi , H. R. Tajik , "Biomedical Image Edge Detection using an Ant Colony Optimization Based on
Artificial Neural Networks", International Journal of Engineering Science and Technology, 2011.
[7] S. Gupta, G. S. Sandhu and N. Mohan, "Implementing Color Image Segmentation Using Biogeography Based
Optimization", Proc. of the International Conference on Computer and Communication Technologies, pp. 167-
170, 2012
[8] V. Zharkova, S. Ipson, J. Aboudarham and B. Bentley, "Survey of image processing techniques", EGSO internal
deliverable, Report number EGSO-5-D1_F03-20021029, October, 2002, 35p. [Online] Available: Temporary
on-line reference link removed - see the PDF document
[9] J. Bozek, M. Mustra, K. Delac, and M. Grgic, "A survey of image processing algorithms in digital
mammography ", Advances in Multimedia Signal Processing and Communications, pp. 631-657, 2009.
[10]S. K. Mahendran and S. S. Baboo, "Enhanced automatic X-ray bone image segmentation using wavelets and
morphological operators", Proc. of the International Conference on Information and Electronics Engineering,
2011.
[11]P. Annangi, S. Thiruvenkadam, A. Raja, H. Xu, X. W. Sun, and L. Mao "A region based active contour method
for X-ray lung segmentation using prior shape and low level features", Proc. of the International Symposium on
Biomedical Imaging, pp. 892- 895, 2010.
[12]F. Ding, W. K. Leow, and T. S. Howe, "Automatic Segmentation of Femur Bones in Anterior-Posterior Pelvis
X-Ray Images", Proc. of the 12th International Conference on Computer Analysis of Images and Patterns, 2007,
pp. 205-212.
[13]N. Senthilkumaran and R. Rajesh, "Edge detection techniques for image segmentation - a survey of soft
computing approaches ", International Journal of Recent Trends in Engineering, 1, pp. 250-255, 2009.

Medical Image segmentation using Image Mining concepts

Medical Image segmentation using Image Mining concepts

More Related Content

What's hot (20)

Similar to Medical Image segmentation using Image Mining concepts (20)

More from Editor IJMTER (20)

Recently uploaded (20)

Medical Image segmentation using Image Mining concepts