Vous êtes sur la page 1sur 6

Proceedings of

International Conference on Emerging Research in Computing, Information,


Communication and Applications
(ERCICA-14)

Analysis of Image Segmentation Techniques for Medical Images


M.S. Abiramia, *, Dr. T. Sheelab
a
Department of Computer Science,Research Scholar,Bharathiar University,Coimbatore,641 046,India
b
Department of IT,Professor & Head,Sai Ram Engg College,Chennai,600 044,India

Abstract

Image segmentation is an image processing technique that divides an image into contiguous regions or segments and typically used to
locate objects and boundaries such as lines, curves, etc. in images. Regions are disjoint with some property for each region such as pixel
intensity, grey level texture, or colour, etc., because a single point cannot be contained in two different regions. Image Segmentation techniques
enable the design of automated segmentation techniques. Several such algorithms are proposed in the literature to simplify and/or change the
representation of an image into something that is more meaningful and easier to analyse. In the proposed system, segmentation of images using
Region Growing algorithm (based on seed pixels) is done in more than one processor. This paper explores several such algorithms which have
been proved to perform on multi-categories of images specifically in medical images. This review paper also suggests interesting directions for
further research.

Keywords: Automated Image Segmentation; Pixels; Thresholding; Region Growing; Edges; Clustering; K-means; FCM; Deformable models.

1. Introduction

Image Segmentation [2] is the division of an image into regions or categories, which correspond to different objects or
parts of objects. Segmentation techniques are based on one of two basic properties of intensity values discontinuity and similarity.
First category is to partition an image based on abrupt changes in intensity, such as edges in an image. Second category is based
on partitioning an image into regions that are similar according to predefined criteria.

Image Segmentation is one of the most important issues in computer aided medical imaging. It is used in the analysis and
diagnosis of numerous applications such as the study of anatomical structure, localization of pathology, treatment planning, and
computer-integrated surgery. Segmentation is a vital role in medical image processing, particularly for abdomen organs
abnormalities detection in Magnetic Resonance Imaging (MRI) and computed Tomography (CT). In medical image processing, an
image is captured, digitized and processed i.e., filtered for segmentation and extracting required information Image segmentation
is still a challenging task for researchers and developers to develop a universal technique for image segmentation. Image
segmentation is also used to differentiate different objects in the image, since our image is divided into foreground and
background, where foreground of image is the region of interest, and background is the rest of the image. Manual segmentation
methods generate errors. These methods are time consuming. Segmentation by expert is variable. Therefore, there is a strong
need to have efficient computer based system that accurately examines the boundaries of abdomen organs with minimum
interaction of user interface.

In this paper a brief study of various automated segmentation techniques [5] are reviewed and discuss advantages and
disadvantages about those techniques. These algorithms differ from each other in aspects like nature of inputs such medical
images, human faces, natural images etc., types of images such as MRI, CT or black and white or colour images etc., and required
accuracy for those images, for example, medical images require more accuracy than natural images.

* Corresponding author. Tel.:+919841075462.


E-mail address: abirami.srm@gmail.com.
1
M.S. Abirami .et.al.

The rest of the paper is divided into four sections. Section 2 discusses various Image Segmentation techniques,
advantages & disadvantages, Section 3 highlights on methodology for abdominal image segmentation techniques and Section 5
deals with topics for further research.

2. Image Segmentation Techniques

Most of the image segmentation techniques are often tailored for specific body regions (brain, abdomen, etc.) and
different image modalities (CT, MRI, etc.). The image segmentation techniques can be classified into four general categories:
pixel-based, region-based, edge-based and model-based techniques. Actually, the basic behaviour of these techniques can be
divided into three concepts. The first concept is the similarity concept like edge-based techniques. Alternatively, the second
concept is based on the discontinuity of pixel values like pixel-based and region-based techniques. Finally, a complete different
approach is the third concept which is based on a statistical approach like Model-based techniques.

2.1. Pixel-based techniques

Using pixel-based techniques each pixel is segmented based on grey-level values, no contextual information, only histogram.
Thresholding [9] is one of the simplest and widely methods of image segmentation. This method is based on a threshold value to
turn a grey-scale image into a binary image (Shapiro, et al., 2001). The most common way to convert a grey level image to a
binary image is to select a single threshold value (T). Then all the grey level values below this T will be classified as black (0),
and those above T will be white (1). The binary image should contain all of the essential information about the position and shape
of the objects of interest (foreground). Threshold technique can be expressed as: T=T[x, y, p(x, y), f(x, y], Where T is the
threshold value. x, y are the coordinates of the threshold value point and p(x,y) ,f(x,y) are points the grey level image pixels.
Threshold image g(x,y) can be defined as: g(x,y) = 1, if f(x,y) > T
0, if f(x,y) T

Threshold segmentation techniques can be grouped into two different classes: global threshold and local (adaptive)
threshold. In the global threshold, a single threshold value is used in the whole image. In the local threshold, a threshold value is
assigned to each pixel to determine whether it belongs to the foreground or the background pixel using local information around
the pixel.

The advantage of obtaining first a binary image is that it reduces the complexity of the data and simplifies the process of
recognition and classification. The main disadvantages are that, in the simplest form only two classes are generated and it cannot
be applied to multichannel images. In thresholding technique, image having only two values either black or white. MR image
contains 0 to 255 grey values.

2.2. Region based techniques

Region Growing technique involves selecting initial seed points and adding neighbouring pixels to the region depending
on the homogeneity criteria. This process is continued until all pixels belong to some region. Thus, region-based segmentation
algorithms [11] operate iteratively by grouping together pixels which are neighbours and have similar values and splitting groups
of pixels which are dissimilar in value.

Algorithm

1. In the first step manually select seed points.


2. In the next steps pixels in the region of seeds are examined and added to the region accordance with the homogeneity
criteria. This process is continued until all pixels belong to some region.
3. In the last step the object illustration is done by growing regions of pixels.

The main disadvantage of this method is, it require user interface for selection of seed points for each region and it is a
time consuming process.

Clustering [1] is the process of organizing objects which are similar and dissimilar objects belonging to other clusters.
Clustering is an unsupervised learning problem. It deals with finding a structure in a collection of unlabelled data i.e., to identify a
finite set of classes known as clusters to classify each pixel. Pixels may belong together on the basis of some specific criterion
such as same colour and/or the same texture and/or distance. Clustering algorithm is classified as Exclusive clustering and
Overlapping clustering. In exclusive clustering, one data (pixel) is belonging to only one cluster then it could not belong to
2
M.S. Abirami .et.al.

another cluster. K-mean is an example of exclusive clustering algorithm. In overlapping clustering, one data (pixel) is belonging
to two or more clusters. Fuzzy C-mean is example of overlapping clustering algorithm.

K-means Clustering [4] MacQueen, 1967 is one of the simplest unsupervised learning algorithms. The procedure is
simple and easy to classify a given data set through a certain number of clusters (assume k clusters) fixed a priori. This algorithm
groups together the objects with similar characteristics to form clusters. The purpose of K-mean clustering is to classify the data.
Also this algorithm aims at minimizing an objective function, in this case a squared error function. The objective function

where is a chosen distance measure between a data point and the cluster centre , is an indicator of the
distance of the n data points from their respective cluster centres.

Algorithm

1. Initially k centroid points are chosen and place them in the space represented by the objects that are being clustered.
2. Assign each object to the group that has the closest centroid.
3. When all objects have been assigned, recalculate the positions of the K centroids.
4. Repeat Steps 2 and 3 until the centroids no longer move. This produces a separation of the objects into groups from
which the metric to be minimized can be calculated.

Fuzzy C-Means Clustering (FCM) [6] In hard clustering (K-means), data is divided into distinct clusters, where each data
element belongs to exactly one cluster. In fuzzy clustering (soft clustering), data elements can belong to more than one cluster,
and associated with each element is a set of membership levels. Fuzzy clustering is a process of assigning these membership
levels, and then using them to assign data elements to one or more clusters. This method (developed by Dunn in 1973 and
improved by Bezdek in 1981) is frequently used in pattern recognition. It is based on minimization of the following objective
function:

where m is any real number greater than 1, uij is the degree of membership of xi in the cluster j, xi is the ith of d-dimensional
measured data, cj is the d-dimension centre of the cluster, and ||*|| is any norm expressing the similarity between any measured
data and the centre. Fuzzy partitioning is carried out through an iterative optimization of the objective function shown above, with
the update of membership uij and the cluster centres cj by:

This iteration will stop when , where is a termination criterion between 0 and 1, whereas k is the
iteration steps. This procedure converges to a local minimum or a saddle point of J m.

Algorithm

1. Initialize U=[uij] matrix, U(0)

3
M.S. Abirami .et.al.

2. At k-step: calculate the centres vectors C(k)=[cj] with U(k)

3. Update U(k) , U(k+1)

4. If || U(k+1) - U(k)||< then STOP; otherwise return to step 2.


The main disadvantage is that it only considers image intensity thereby producing unsatisfactory results in noisy images.

2.3. Edge based techniques

Edge based methods [7] attempt to solve the image segmentation by detecting the edge between different regions. This
technique determines the presence of an edge or line in an image and outlines them in an appropriate way. The main purpose of
edge detection is to simplify the image data in order to minimize the amount of data to be processed. Generally, an edge is defined
as the boundary pixels that connect two separate regions. The detection operation begins with the examination of the local
discontinuity at each pixel element in an image. Based on some essential characteristics such as amplitude, orientation, and
location of a particular subarea in the image, the detector has to decide whether each of the examined pixels is an edge or not.

The edge based method has the advantage that it analysis the images by drastically reducing the amount of data to be
processed, and also preserving useful structural information about object boundaries. With only edge detector algorithms high
quality cannot be achieved for ultrasound images which have inherent speckle noise and texture characteristics. Due to this
drawback edge based method rarely used alone.

Hybrid image segmentation is a combination of edge based and region based techniques. In this image is firstly
partitioned into regions and then merged using split and merge technique and after that it detects the contours using edge-based
technique.

2.4. Model based techniques

Deformable models are curves or surfaces, for segmentation in the image domain, or hyper-surfaces, for the segmentation of
higher dimensional image data, such as stacks of images, which deform under the influence of internal and external forces to
delineate object boundary. The internal forces are defined such that they preserve the shape smoothness of the model, while the
external forces are defined by the image features to drive the model toward the desired position i.e., to the desired region
boundaries. It is difficult for parametric deformable models to adapt the model topology during deformation. However geometric
deformable models are designed to handle topological changes.

Two general classes of deformable models are the parametric deformable models and the geometric deformable models.
Parametric deformable models referred as snakes or active contour models [10] are represented explicitly as parameterized curves
in a Lagrangian formulation. Geometric deformable models are represented implicitly as level sets of two-dimensional distance
functions that evolve according to an Eulerian formulation. The deformable model that has attracted the most attention to date is
popularly known as snakes (Kass et al., 1988). Deformable curve, surface and solid models gained popularity after they were
proposed for use in computer vision (Terzopoulos et al., 1988) and computer graphics (Terzopoulos and Fleischer, 1988).

A classic snake model is described by x(s) = (x(s), y(s)), where s is the arc length, and x(s) and y(s) are x and y coordinates
along the contour, and the energy of the model is given by E snake=snake E(x(s))ds = snake Eint(x(s) + Eext(x(s))ds, where Eext is the
external energy, and Eint is the internal energy, given by
Eint = ((t) x(s)/ s2 + (t) x(s)/ s2 ),
4
M.S. Abirami .et.al.

where and are the coefficients that control the snakes tension and rigidity, respectively. The goal is to find a snake, x(s), that
minimizes Esnake. The external energy is in accord with the image features, and for a given image f(x, y),
Eext = - {G(x,y) * f(x,y)},
where G(x,y) is the two-dimensional Gaussian kernel with as the standard deviation.

The classical active contour models have several limitations. One limitation is that the initial contour must be close to
the true boundary or it will likely converge to a wrong result. To address this problem, increase the capture range of the external
force fields and guide the contour toward the desired boundary. Another limitation is the poor convergence of classical snakes to
boundary concavities. The GVF snake is an effective model that can be employed to solve this problem.

The second class of deformable models, namely the geometric models, use a distance transformation to define the shape
from the n-dimensional to an n + 1-dimensional domain, where n 1 for curves, n 2 for surfaces on the image plane, etc. Thus,
it uses the level-set based shape representation, transforming the curves into higher dimensional scalar functions.
geometric deformable models were introduced independently by Malladi et al. and Caselles et al..

The level set method views a moving curve as the zero level set of a higher dimensional function (x, t). Generally, the
level set function satisfies,
(x,t) < 0 in (t),
(x,t) = 0 in C(t),
(x,t) > 0 in Rn\ (t),
where the artificial time t denotes the evolution process, C(t) is the moving curve, and (t) represents the region possibly multi-
connected) that C(t) encloses. An evolution equation for the curve C moving with speed F in its normal direction is
given by
t = F(x)
Here, the surface = 0 corresponding to the propagating hyper surface may change topology, as well as form sharp
corners. A particular case is motion by mean curvature, when F = div((x)/| (x)|) is the curvature of the level curve of
passing through x. The above equation becomes
/t = . Div ( / ), with (0, x) = 0(x) and t (0,).

The main advantage of parametric models is that they are robust and usually very fast in their convergence, depending
on the predetermined number of control points. However, an obvious weakness of these models is that they are topology
dependent: a model can only capture a single ROI, and therefore, in images with multiple ROIs, multiple models have to be
initialized, one for each ROI.

3. Proposed Methodology

The proposed methodology is to develop an automated hybrid segmentation technique for segmenting abdominal medical
images and implement it in parallel, and then calculating the different features of those segmented regions. Thus the objective of
the proposed methodology is to improve the performance in the execution time (segmentation and feature extraction) of the
images.

4. Conclusion

Image Segmentation algorithms offer an attractive approach to automated design of segmentation for medical applications.
They eliminate the problems in manual segmentation, and provide a mean for constructing even hybrid segmentation techniques.
In this paper, we have focused on a family of such algorithms that construct automated image segmentation methods that deal
with multi-categories especially in medical images.
The potential areas for further research would be
Developing a hybrid segmentation scheme that performs segmentation of medical images to yield superior performance.
Optimization of algorithms by parallel segmentation to reduce its segmentation time.

5
M.S. Abirami .et.al.

References

1. P.TamijeSelvy, V. Palanisamy, T.Purusothaman. Performance Analysis of Clustering Algorithms in Brain Tumor Detection of MR Images. European
Journal of Scientific Research, Vol.63, No.3, 2011.
2. M.A.Balafar, A.R.Ramli, M.I.Saripan, S.Mashohor. Review of brain MRI image segmentation methods. Artif Intell, Springer Science, DOI:10.1007/s10462-
010-9155-0. Rev 2010.
3. M. Masroor Ahmed, Dzulkifli Bin Mohamad. Segmentation of Brain Magnetic Resonance Images (MRIs): A Review, Int. J. Advance. Soft Comput. Appl.,
Vol.3, No. 3, ISSN 2074-8523, November 2011.
4. Y H.P. Ng, S.H. Ong, K.W.C. Foong, P.S. Goh, W.L. Nowinski. Medical Image Segmentation using K-means Clustering and Improved Watershed Algorithm.
IEEE Xplore, 2009.
5. Daw-Tung Lin, Chung-Chih Lei and Siu-Wan Hung. Computer-Aided Kidney segmentation on Abdominal CT Images. IEEE Transactions on Information
Technology in Biomedicine, vol. 10, no. 1, pp. 59 - 65, January 2006.
6. J. Selvakumar, A. Lakshmi, T. Arivoli. Brain tumor segmentation and its Area calculation in Brain MR images using K-Mean clustering and Fuzzy C-Mean
Algorithm. IEEE International Conference on Advances in Engineering Science and Management, March 2012.
7. Hu A, Grossberg B, Mageras C. Survey of recent volumetric medical image segmentation techniques. Biomedical Engineering. 2009;321-346.
8. Zuva T, Olugbara OO, Ojo SO et al. Image segmentation, available techniques, developments and open issues. Canadian Journal on Image Processing and
Computer Vision 2011;2(3):20-29.
9. Orlando J. Tobias, Rui Seara. Image Segmentation by Histogram Thresholding Using Fuzzy Sets. IEEE Transactions on Image Processing. Vol. 11, No. 12,
December 2002.
10 . Muralidhar G, Bovik A, Giese J et al. Snakules: A model-based active contour algorithm for the annotation of spicules on mammography. IEEE Transactions
on Medical Imaging. 2010;29(10):1768-1780.
11. Benjamin Wah. Image Segmentation by Region Based and Watershed Algorithms. Computer Science and Engineering, 2008.
12. Mrs. Bharati R. Jipkate, Dr. Mrs. V. V. Gohokar, A Comparative Analysis of Fuzzy C-means Clustering and K-means Clustering Algorithms. International
Journal of computational Engineering Research, Vol.2, Issue 3, May-June 2012.
13. Heimann T, Meinzer HP, Wolf I. A statistical deformable model for the segmentation of liver ct volumes. MICCAI Workshop on 3d Segmentation in the
Clinic: A Grand Challenge (2007), 161-166.
14. Kai Xiao, Sooi Hock Ho, Andrzej Bargiela. Automatic brain MRI segmentation scheme based on feature weighting factors selection on fuzzy c-means
clustering algorithms with Gaussian smoothing. Int. J. Comput. Intelligence in Bioinformatics and Systems Biology,Vol. 1, No. 3, 2010.
15. Tara.Saikumar. MRI Brain Image Segmentation Algorithm Using Watershed Transform and Kernel Fuzzy C-Means Clustering on Level Set Method.
International Journal on Computer Science and Engineering (IJCSE), ISSN : 0975-3397 Vol. 3 No. 4 Apr 2011.
16. Tsechpenakis G, Metaxas D (2009) CoCRF deformable model: a geometric model driven by collaborative conditional random fields. IEEE Trans Image
Process 18(10):23162329.
17. Pham DL, Xu C, Prince JL. A survey of current methods in medical image segmentation. Annual Review of Biomedical Engineering, 2000;2:315-338.
18. Sanjay Saxena, Neeraj Sharma and Shiru Sharma. Image Processing Tasks using Parallel Computing in Multi core Architecture and its Applications in
Medical Imaging. International Journal of Advanced Research in Computer and Communication Engineering, Vol. 2, Issue 4, April 2013.
19. Vishal B. Padole, D. S. Chaudhari. A Review of Segmentation Methods for Detection of Brain Tumor in MRI. International journal of electronics
communication & soft computing science and engineering (IJECSCSE), Vol. 1, Issue 1, 2011.
20. GauriAnandgaonkar, PoojaPatil, Prof .G. S. Sable. An overview: Image segmentation Techniques for Brain Images, International Journal of Computer
Technology and Electronics Engineering (IJCTEE), Vol.3, Special Issue March-April 2013
21. Xinjian Chen, Jayaram K. Udupa, Ulas Bagci. Medical Image Segmentation by Combining Graph Cuts and Oriented Active Appearance Models. IEEE
Transactions on Image Processing. Vol. 21, NO. 4, Apr; 2012.
22. Jay Patel, Kaushal Doshi. A Study of Segmentation Methods for Detection of Tumor in Brain MRI. Advance in Electronic and Electric Engineering.
ISSN 2231-1297, Volume 4, Number 3 (2014), pp. 279-284
23. Muhammad Waseem Khan. A Survey: Image Segmentation Techniques. International Journal of Future Computer and Communication, Vol. 3, No. 2,
April 2014.
24. Lin Li, Peeter Ross, Maarja Kruusmaa, Xiaosong Zheng. A Comparative Study of Ultrasound Image Segmentation Algorithms for Segmenting Kidney
Tumors. ISABEL '11, October 26-29, Barcelona, Spain.
25. El-Baz, Acharya, Mirmehdi, Suri. Multi Modality State-of-the-Art Medical Image Segmentation and Registration Methodologies: Volume 1,
DOI 10.1007/978-1-4419-8195-0_2, # Springer Science+Business Media, LLC 2011.
6

Vous aimerez peut-être aussi