Brain Tumor Segmentation by EM Algorithm

 

Swarnakala¹, Natarajah Srikumaran²

¹Research Scholar, Department of Marine Biotechnology, AMET University, Chennai

²Department of Marine Biotechnology, AMET University, Chennai

 

ABSTRACT:

Brain tumor segmentation system for medical images is provoked by the requirement of high accuracy when dealing with human life. In this paper, an automatic system for the identification and segmentation of MRI brain tumor images by using the Expectation Maximization (EM) segmentation algorithm is proposed. Here the MRI brain images are given as the input for the system and the segmentation of normal and abnormal tumor images can be obtained. The received EM based tumor images can be used for image classification also. This segmentation process gives more accurate results than the other image segmentation algorithms.

 

 

 

INTRODUCTION:

Based on the central brain tumor registry statistics of United States, the brain tumor is considered as the leading causes of death by cancer around the world. For the identification and segmentation of brain tumor from the MRI brain images, there are many segmentation algorithms, and some of them are stated below. 

 

An improved description of the watershed and Fuzzy C-Means (FCM) clustering algorithm is discussed in¹. The drainage method is done to prevent the over segmentation and FCM is done for histogram based centroid selection. The idea of semantic feature layer based feature is discussed². PCA does the feature extraction, and SFL algorithm and the classification is done by PNN classifier.

 

A method for segmentation of Alzheimer’s disease in MRI brain images is discussed³. The texture features are extracted from the brain regions and are classified by using the random forest classifier.

 

A method to characterize the MRI brain image by extracting the features using FCM algorithm and is classified by using the SVM classifier into tumor affected and tumor free regions is discussed⁴.

 

What a multi-level brain partitioning for extraction of SIFT features from brain images is discussed⁵. The extracted features are classified y using the SVM classifier. A brain tumor segmentation system is discussed in⁶. First, the image acquisition is made followed by the extraction of features by FCM clustering algorithm. An acoustic study of atropine sulfate present in the water of various concentrations at 35°C using ultrasonic interferometer is discussed in⁷. A study of the molecular interaction of pharmaceutical drug of 2- acetic acid in water of various concentrations at 303K is presented in⁸. A synthesis study of hydrazine-based novel HMG based CoA inhibitor and its docking studies are explained in⁹. A controllability of the impulsive neutral functional of the integro-differential inclusions with an infinite delay and is described in¹⁰.

 

PROPOSED METHODOLOGY:

The proposed methodology explains about the system and its working of what we have done for the brain tumor segmentation process of the tumor detection from MRI brain image. The block diagram of proposed system is shown in fig 1.  

 

Figure 1. Block diagram of the Proposed Segmentation System

 

PRE-PROCESSING:

The image segmentation process starts with the pre-processing step where the input images are taken and is pre-processed by denoising the MRI brain images and the also color conversion process occurs where the color images are changed into gray scale images. 

 

EM SEGMENTATION:

The next step is the segmentation process where the tumor is segmented from the original images. In this step, the pre-processed images are given as the input to the segmentation process in which the EM algorithm finds out the locally maximum likelihood parameters from a statistical model of the images. The EM iteration changes between a prospects steps that create a task for the expectation of log likelihood calculation using the parameter estimations. The segmentation results will be obtained either as a standard tumor images or the abnormal tumor images.

 

RESULTS AND DISCUSSION:

The results that are obtained for our proposed brain tumor segmentation process using the EM segmentation algorithm is shown below.

 

(a)

 

(b)

 

(c)

 

(d)

 

(e)

 

(f)

 

Fig.4.1. (a) and (b) Pre-Processing Output (c), (d), (e), (f) Step by step EM segmentation output

 

From the above image of the segmentation output images of the proposed method, it is understood how the segmentation process of the tumor cell regions is extracted from the original images as shown.

 

CONCLUSION:

From the above images we can conclude that out proposed system of brain tumor segmentation based on the EM based algorithm helps us to segment the exact regions of the tumor cells that are present in the MRI brain images. The inputs images are taken from the datasets that are used widely in much brain tumor segmentation system. The proposed EM segmentation algorithm gives more accurate segmentation results than other methods. In future, we can use this segmented output images for the classification of the brain tumor cells.

 

REFERENCE:

1.     Benson, C. C, Deepa, V., Rajesh, V. L., and Rajamani, K. “Brain tumor segmentation from MR brain images using improved fuzzy c-means clustering and watershed algorithm”. IEEE International Conference on Advances in Computing, Communications and Informatics, pp. 187-192, 2016.

2.     Bhuvaneswari, K. S., and Geetha, P. “Semantic feature based classification of Brain MRI using PCA and PNN”. IEEE International Conference on Electrical, Electronics, and Optimization Techniques, pp. 2700-2706, 2016.

3.     Chaddad, A., Desrosiers, C., and Toews, M. “Local discriminative characterization of MRI for Alzheimer's disease”. IEEE 13^th International Symposium on Biomedical Imaging, pp. 1-5, 2016.

4.     Halder, A., and Dobe, O. “Detection of tumor in brain MRI using fuzzy feature selection and support vector machine”. IEEE International Conference on Advances in Computing, Communications and Informatics, pp. 1919-1923, 2016.

5.     Li, T., and Zhang, W. “Classification of brain disease from magnetic resonance images based on multi-level brain partitions”. IEEE 38^th Annual International Conference of the Engineering in Medicine and Biology Society, pp. 5933-5936, 2016.

6.     Sehgal, A., Goel, S., Mangipudi, P., Mehra, A., and Tyagi, D. “Automatic brain tumor segmentation and extraction in MR images”. IEEE Conference on Advances in Signal Processing, pp. 104-107, 2016.

7.     Balakrishnan, J., Balasubramanian, V., Rajesh, S., Sivakumar, M. “Acoustic study of atropine sulphate in water of various concentrations at 35°C using ultrasonic interferometer”, Journal of Chemical and Pharmaceutical Research, vol. 4, no. 9, pp. 4283-4288,2012.

8.     Rajesh, S., Balasubramanian, V. “Molecular interaction studies on pharmaceutical drug of 2-[1-(aminomethyl) cyclohexyl] acetic acid in water of various concentrations at 303K”, Journal of Chemical and Pharmaceutical Research vol. 5, no. 2, pp. 131-138, 2013.

9.     Saravanan, B., Saravanan, R. R., Manivannan, V. “Synthesis of hydrazine based novel HMG coA inhibitor and its docking studies”, International Journal of Drug Development and Research, vol. 5, no. 4, pp. 293-299, 2013.

10.   Manimaran, S., Gunasekar, T., and Subramaniyan, G. V. “Controllability of impulsive neutral functional integro differential inclusions with an infinite delay”. Global Journal of Pure and Applied Mathematics, vol. 10, no. 6, pp. 817-834, 2014.

.

 

 

 

Accepted on 05.09.2017        © RJPT All right reserved