Nagesh Vadaparthi, Srinivas Yarramalle, Suvarna Kumar G, Vamsee Krishna V/ International Journal of

Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com

Vol. 2, Issue 2,Mar-Apr 2012, pp.666-668

AN IMPROVED MEDICAL IMAGE SEGMENTATION USING

CHARGED FLUID MODEL

Nagesh Vadaparthi*1, Srinivas Yarramalle2 , Suvarna Kumar G3,

Vamsee Krishna V4
1
Department of I.T, M.V.G.R. College of Engineering, Vizianagaram, Andhra Pradesh, India.
2
Department of I.T, GITAM University, Visakhapatnam, Andhra Pradesh, India.
3
Department of C.S.E., M.V.G.R. College of Engineering, Vizianagaram, Andhra Pradesh, India.
4
Department of C.S.E., M.V.G.R. College of Engineering, Vizianagaram, Andhra Pradesh, India.
Abstract
Segmentation has a major role to play in the field of medical imaging for effectively detecting the deformities. Though
there are huge medical imaging techniques available, Magnetic Resonance Image (MRI) is widely used because of the
non-ionizing radiation which is being used. In this paper, we proposed a refined method using charged fluid model
(CFM) which yields better results. It is essential to use an effective algorithm which can automatically segment the
region of interest. Hence, we used Otsu’s algorithm to segment the image automatically.

Keywords:- Charged Fluid Model, Segmentation, magnetic resonance images, region of interest.
technique and knowledge based approaches [9-11].
INTRODUCTION Among these, deformable models are the most popular
Medical imaging is the technique and process used to models due to their ability to recover the shape of
create images of the human body (or parts) for clinical biological structures much more accurately in various
purposes or medical science. Medical imaging uses state- segmentation applications [12-13]. The deformable
of-the-art technology to provide 2 or 3-dimensional models can be broadly classified into parametric and
images of the living body. Imaging studies can diagnose geometric models. A moving equation should be defined
disease or dysfunction from outside the body, providing to drive the initial contours to the structure boundaries
information without exploratory surgery or other invasive due to which these algorithms are viewed as a modeling
and possibly dangerous diagnostic techniques. In recent of curve evolution [14]. There are various algorithms
years, the field of medical imaging has undergone drastic proposed in these deformable models among which
changes which in turn have facilitated the physicians in Charged Fluid Model (CFM) [15] is much accurate. But
effectively diagnosing the diseases. But, still there is lot there are few limitations existing in this algorithm like,
of scope to derive new techniques for effectively this algorithm requires initialization inside the region of
identifying the disease. In the field of medical imaging, interest and the algorithm also needs some improvisations
segmentation plays a major role in taking the pre-surgery for interactive segmentation. Hence to improve these
and post-surgery decision for faster recovery of the drawbacks, in this paper we propose an additional
diseases. algorithm named Otsu’s algorithm [16] which is used to
automatically perform histogram shape-based image
Segmentation is a process of partitioning an image into
thresholding [17].
non-intersecting regions which have two properties such
as homogeneity within a region and heterogeneity The paper is organized as follows: Section-II
between the regions [1].There are various algorithms describes about the segmentation algorithm of CFM and
available for image segmentation process [2–6]. But the section-III explains the experimental results of the
most widely used segmentation algorithm specifically for proposed method. Section-IV demonstrates the evaluation
medical images is based on gray-level value of the pixel of the proposed method and finally, section-V concludes
as the medical images are of gray-level only. There are the paper.
various medical imaging techniques such as CT, MRI,
and PET etc. Among these, MRI is widely used medical SEGMENTATION ALGORITHM
imaging technique because of the non-ionizing radiation Physics-based probabilistic systems were being used to
property that is used and having an excellent contrast investigate and analyze the biological models. In charged
between the soft tissues [7]. The MRI data of the patient fluid models, each fluid element has its own charge as if
is collected and is stored in the form of an image. In this it has been calculated by interpolating the charges of the
MRI analysis, the difficult task is segmenting the region covered particles [15]. As there are few drawbacks in the
of interest. To achieve this, various algorithms have been CFM model stated in [15] like there should be an
proposed in the literature. But, these segmenting effective technique which can automatically segment the
algorithms differ from application to application. There is image, in this paper we introduce a new approach by
no specific algorithm available which serves for all including Otsu’s algorithm to effectively segment the
applications [8]. The various segmentation algorithm image for extracting the region of interest which in turn
proposed earlier are threshold-based, region-based, can help the medical practitioners to identify the
statistics-based, deformable models, atlas guided
680 | P a g e
 Nagesh Vadaparthi*1, Srinivas Yarramalle2 , Suvarna Kumar G3, Vamsee Krishna V/ International
Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 2, Issue 2,Mar-Apr 2012, pp.680-682

deformity and its density for pre-surgery and post-surgery Algorithm 4: Extraction of region of interest
treatment. In this technique, the Otsu is used to _____________________________________________
automatically perform the histogram shape-based image
thresholding. The advantage of using Otsu algorithm is 1) Compute histogram and probabilities of each intensity
that, the set of thresholds is selected automatically and level
stably based on integration of the histogram. The 2) Set up initial ωi(0) and μi (0).
segmentation of the medical image involves 3 algorithms 3) Step through all possible thresholds maximum
which are demonstrated below and the numbers specified intensity
in the brackets refer to the equations derived in paper [15]. 1. Update ωi and μi
______________________________________ 2. Compute σb2(t)
4) Desired threshold corresponds to the maximum σb2(t)
Algorithm 1: Charged fluid (Ref.-15)
______________________________________
RESULTS & DISCUSSION
1) parameter setting of in (11)* The experimentation of the proposed algorithm
2) image potential computation using (11)* has been implemented on various images of brain and
breast. All the algorithms have been implemented on a
3) repeat(i) personal computer with CPU of Core i3 Processor
2.4GHz using MATLAB. In order to carry out the
a) uniform charge distribution over fluid proposed work, we have considered the data sets of the
elements brain images from brainweb images [18]. The advantages
b) repeat(j) of using the brainweb database are that the comparision
Algorithm 2 can be made available using the ground truth for the
c) until(j) electrostatic equilibrium is achieved tissue classes (CSF, WM and GM) from which the digital
by setting in (10)* phantom were created [19].
d) 1–pixel-wide front construction using the
boundary element detection method
e) Algorithm 3 The original images are shown in Figure 1(a) breast
f) mean potential computation and charge image and 1(b) brain image having deformities. The
normalization using (5)* output results are shown in figure 2(a) Breast image and
2(b) Brain image after segmenting and extracting the
4) until(i) no deformation in the charged fluid shape region of interest.
5) Algorithm 4 for extracting the region of interest

___________________________________________

Algorithm 2: Charge distribution procedure (Ref.-15)

____________________________________________

1) forward FFT computation of the charge array based

upon (16)*
2) inverse FFT computation of the potential array based
upon (17)*
3) electric field computation using (7)* a) Breast Image b) Brain Image
4) advance distance computation using (9)* Figure 1: Original images having deformities
5) charge density computation using the SUDS based
upon (15)*
____________________________________________

Algorithm 3: Front deformation procedure (Ref.-15)

____________________________________________

1) mean electric field compensation using (18)*

2) effective field computation using (12)*
3) 2–pixel-wide interface localization based upon Fig. 6
in [15]

*indicates the equations used from reference paper [15]. a) Breast Image b) Brain Image
__________________________________________ Figure 2: Output images having showing deformities

681 | P a g e
 Nagesh Vadaparthi*1, Srinivas Yarramalle2 , Suvarna Kumar G3, Vamsee Krishna V/ International
Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 2, Issue 2,Mar-Apr 2012, pp.680-682

The proposed method is implemented using MATLAB [12] T. McInerney and D. Terzopoulos, “Deformable
software. The experimentation is carried out using Core models in medical image analysis: A Survey”, Med.
i3 processor and the time taken to execute, this method is Image Anal., vol.1, no. 2, pp. 91-108, 1996.
about 0.326 seconds when compared to the existing [13] S. Osher and N. Paragios, Geometric Level Set
method which consumes approximately about 0.51 Methods in Imaging, Vision and Graphichs”, New
seconds. York: Springer-Verlag, 2003.
[14] Zhen Ma, Joao Manuel R. S. Tavares and R. M.
CONCLUSION Natal Jorge, “ A Review on the Current
The segmentation has a vital role in the field of medical Segmentation Algorithms for Medical Images”,
imaging for effective diagnosis of the patient having White Paper.
deformities help the medical practitioner in performing [15] Herng-Hua Chang, Daniel J. Valentino, Gary R.
the pre-surgery and post-surgery process. Hence in this Duckwiler and Arthur W. Toga, “ Segmentation of
paper, we proposed an effective segmentation method Brain MR Images Using a Charged Fluid Model”,
based on CFM model to segment the medical image. IEEE Transactions on Biomedical Engineering, Vol.
Hence, our proposed method yields better results when 54, no. 10, Oct. 2007.
compared to the existing method and is much faster when [16] N. Otsu, “A Threshold Selection Method from
compared to the existing model. Gray – Level Histogram”, IEEE Transaction
Systems Man. And Cybernetics, Vol. 9, pp. 62-66,
REFERENCES 1979.
[1] R.C Gonzalez and R. E Woods, “Digital Image [17] M. Sezgin and B. Sankur, “Survey over image
Processing”, Addison-Wesley, 1992. thresholding techniques and quantitative
[2] D. L. Pham, C. Y. Xu, and J. L. Prince, “A survey performance evaluation”, Journal of Electronic
of current methods in medical image segmentation”, Imaging 13(1): 146-165, 2003.
Annu. Rev. Biomed. Eng. vol. 2, pp. 315–337 [18] BrainWeb: Simulated Brain Database [Online],
(2000). Available: http://www.bic.mni.mcgill.ca/brainweb
[3] K. Van Leemput, F. Maes, D. Vandeurmeulen, and [19] OnTsang et al, “Comparison of Tissue
P. Suetens, “Automated model-based tissue Segmentation algorithms in NeuroImage analysis
classification of MR images of the brain”, IEEE software tools”, 30th Annual International IEEE
Trans. Med. Imag. vol. 18, no. 10, pp. 897–908 (Oct. EMBS Conference, Vancouver, British Columbia,
1999). Canada, pp.3924-3928, August 20-24, 2008.
[4] G. Dugas-Phocion, M. Á. González Ballester, G.
Malandain, C. Lebrun, and N. Ayache, “Improved
EM-based tissue segmentation and partial volume
effect quantification in multi-sequence brain MRI”,
Int. Conf. Med. Image Comput. Comput. Assist. Int.
(MICCAI). pp. 26–33, (2004)
[5] K. Van Leemput, F. Maes, D. Vandermeulen, and P.
Suetens, “A unifying framework for partial volume
segmentation of brain MR images”, IEEE Trans.
Med. Imag. vol. 22, no. 1. pp. 105–119 (Jan. 2003).
[6] M. Prastawa, E. Bullitt, S. Ho, and G. Gerig,
“Robust estimation for brain tumor segmentation”,
Int. Conf. Med. Image Comput. Comput. Assist.
Inter (MICCAI). pp. 530–537 (2003).
[7] Nagesh Vadaparthi, Srinivas Yarramalle and Suresh
Varma, “Unsupervised Medical Image
Segmentation on Brain MRI Images using Skew
Gaussian Distribution”, IEEE-International
Conference on Recent Trends in Information
technology, pp.1293-1297, June 2011.
[8] N. R. Pal and S. K. Pal, “A review on image
segmentation techniques”, Patt. Rec. 26:1277-1294,
1993.
[9] I. N. Bankman, “Handbook of Medical Imaging”,
San Diego, CA: Academic, 2000.
[10] J. K. Udupa and G. T. Herman, “3D Imaging in
Medicine”, 2nd ed., Boca Raton, FL: CRC, 2000.
[11] A. P. Dhawan, “Medical Image Analysis”, New
York: Wiley, 2003.

682 | P a g e