GRD Journals- Global Research and Development Journal for Engineering | Volume 2 | Issue 4 | March 2017

ISSN: 2455-5703

Brain Tumour Extration From MR Images

Using Segmentation Techniques: A Review
Ankita Kapil Dr. (Mrs.) Shailja Shukla
Student Professor
Jabalpur Engineering College, Jabalpur, MP India Jabalpur Engineering College, Jabalpur, MP India

Abstract
Medical image processing has dramatically revolutionized the health care sector by helping clinicians towards early and accurate
diagnosis of the disease. MRI is one of the most versatile and widely used imaging modality. Brain tumours are the abnormal
growth of tissues and this can be benign and malign subject to the tumour location and size, which are often difficult to visually
diagnose from MRI film. Therefore, several image processing techniques have been developed and in use for accurate and early
detection. Image Segmentation typically applied to detect a specific region in an image and is often used in biomedical field. One
of the most challenging implementation of this technique is in brain tumour recognition. This paper presents a comparative study
of different segmentation techniques for extraction of tumour from MRI images.
Keywords- Image processing, brain tumour, Image segmentation, Magnetic resonance imaging (MRI)

I. INTRODUCTION
Medical image processing is a challenging and a crucial field, the analysis of the taken medical image is very important task for
further diagnosis of the disease in patients. There are several techniques that help in the deformed object recognition in these
images [1][2][3].A brain tumour is a mass of abnormal tissues or cells in the brain. A tumour can be benign or malignant. Benign
tumours are the grade 1 or 2 type of tumours, they are not cancerous and do not grow back after treatment and diagnose. The
malignant tumours are grade 3 or 4 type of tumours, which contains cancerous cells, malignant tumours grow rapidly and spreads
in brain and surrounds the brain tissues, there are over 120 types of tumours depending on their generation part, size, origin of cells
from they are growing and other causes[3]. Different type of tumours affects different part of brain which may cause in person
behavior, health, and senses. The diagnose can be carried out by knowledge of medical imaging [4], MRI, CT Scans,
Electroencephalography (EEG). MRI has advantages over CT scans being capable of detecting the flowing blood and cryptic
vascular malformation. MRI is able to visualize the brain in all three planes Axial, Sagittal, Coronal which gives a good view of
brain as given below in figure.

Fig. 1: Three axis views of brain (axial, sagittal, coronal)

For brain tumour identification usually the T1- T2 weighted images are used and they are differentiated by looking at their
csf (cerebrospinal fluid), in T1 weighted images csf is dark and bright in T2 weighted images. Contrast type T1 images are
obtained by applying gadolinium, it makes the image bright. MRI images are used widely to detect the tumours in brain, for noise
removal in different images filtering is applied and for different methods an analysis is shown here in [5][6], there are many
methods performed on MRI images like segmentation, morphological operations, feature selection and extraction, classification
etc.

All rights reserved by www.grdjournals.com 14

 Brain Tumour Extration From MR Images Using Segmentation Techniques: A Review
(GRDJE/ Volume 2 / Issue 4 / 003)

II. LITERATURE REVIEW

Segmentation of MR images divides the regions containing pixels with same attributes for the meaningful and useful analysis.
Multi-channel MR images with varying contrast characteristics can be acquired, providing additional information for Image
Segmentation for distinguishing between different structures. For general reviews on the segmentation of MR images, see [7-10].
The techniques of segmentation can be generally classified as Thresholding, Region growing, Classifier, Clustering, Markov
random analysis, ANN, Atlas guided approach. In this paper some techniques are discussed a reviewed.

A. Thresholding
Thresholding in image segmentation converts a grayscale image into a binary image, it divides the different attributes in an image
between foreground and back ground, say an image is converted into a black(0 value pixel), and white(1 value
pixel).If g(x, y)is a thresholded version of f(x, y)at some global threshold T,
1 if f(x, y) > 0
g(x) = {
0 otherwise
Morphological operations (Erosion, Dilation) are applied on the image obtained after Thresholding. Below the image shows
global Thresholding on brain tumour MRI.

Fig. 2: a) brain MRI b) after global Thresholding c) final tumour detection after morphology

In [11]Natarajan Prem et al.represented detection of brain tumour by MRI images using pre-processing where image
sharpening is performed by median filters image enhancement by histogram , segmentation by Thresholding and morphological
operations , and finally detection of tumour by image subtraction. As result the tumour is detected by Thresholding and the
operations performed and the time taken not considered here. In EEG also Thresholding removes artifacts and gives better result
Roy, Vandana, and Shailja Shukla [12].

B. Region Growing Based

The goal of region growing is to use image characteristics to map individual pixels in an input image to sets of pixels called regions.
The region corresponds to a world object or a meaningful part of one. When the pixels are arranged on the basis of their properties
and neighbouring pixel properties its local technique, when pixels are grouped and formed a big region from whole image it is a
global technique, and in splitting and merging an individual pixels or set of pixels below the figure shows the example of region
growing based segmentation.

Fig. 3: diffusion- weighted brain image (left) result of region growing segmentation (right)

All rights reserved by www.grdjournals.com 15

 Brain Tumour Extration From MR Images Using Segmentation Techniques: A Review
(GRDJE/ Volume 2 / Issue 4 / 003)

In Zabir, Ishmam, et al. [13] the region growing is automatically selected as initial contour to the iterative distance
regularized level set evolution method so the need of selection of the region of interest by the user. A completely computer aided
automated technique is developed to detect glioma the BRATS 2012 database is used which contains different glioma cases. T2
and flair images are used and final contour after 50 iterations both the images are summed up to find final segmented tumour, the
rese performance was 55-95% dice coefficient for different cases.

C. Clustering Based
A cluster consist a group of similar objects, clustering is of two types supervised and unsupervised, in supervised clustering criteria
is decided by user but in unsupervised clustering criteria is decided by the system itself.

D. K- Means Clustering
K-Means is an unsupervised clustering algorithm that classifies the input data points into multiple classes based on their inherent
distance from each other. Let X ={x1, x2, x3xn} represents a set of elements of an image and C= {c1, c2, c3ck) represents
asset of centroid, where is number of clusters. The main objective of k means clustering is to minimize an objective function J that
determines the closeness between the pixel and the cluster centroids, and is calculated as follows,
J= =0 =0 2
CC Lin, CC Chang. [14], in their work k - means converts the grayscale image into a colour space image and then the
position of tumour is separated from the other objects in the MR image, first the grayscale image converted into RGB colour and
then further converted into CIELab colour model, they concluded that the method is easy and efficient at that time.

E. Fuzzy C- Means Clustering

fuzzy c means clustering allows the objects to belong to one or more clusters, in FCM algorithm number of elements partitioned
into c fuzzy clusters with given range or criteria. Chuang, Keh-Shih, et al. [15], applied the FCM algorithm that includes the spatial
information into the membership function to fully discover the characteristics of the image, the advantage of doing this yields the
region homogenous than the other methods, removes the noisy spots, the result showed that the noise was less with the new
algorithm in comparison with the conventional and Benson, C. C., et al. [16] represented a method to segment brain tumour where
the FCM and watershed together is applied. FCM clustering is applied for initial centroid selection based on histogram calculation
and in watershed an atlas based marker detection method was applied to reduce over segmentation, the result found here was
accurate from 93.13 and 88.64 % of dice and tanimoto coefficient values.

F. Watershed
watershed is another region based method that has its origin in mathematical morphology, general concept was introduced by H.
Digabel and C. Lantujoul in 1977,[17] watershed segmentation is based on the idea of regarding an image as topographic
landscape with ridges and valleys. K. Bhima and A. Jagan [18], this paper concentrate on improving the detection of brain tumour
in MRI by using marker based watershed segmentation in few dataset MRI images, followed by MATLAB implementation as
testing results they found better results, the brain tumour was clearly segmented. I. Maiti and M. Chakraborty,[19], brain tumour
detection based on colour by using watershed along with edge detection algorithm, the result found was promising, it was based
on HSV colour madal.P. Dhage, M. R. Phegade and S. K. Shah [20], the paper proposed a method based on watershed segmentation
that separates the abnormal tissues of brain from the normal one, while the tumour is extracted but the position and shape is also
determined and other parameters like eccentricity, entropy and centroid is also calculated.

G. Atlas Based Segmentation

In atlas based segmentation data obtained from different subjects arranged as common anatomy an atlas (brain images) is
constructed.SRI24 is an atlas of 24 normal brain subjects which is given by T. Rohlfing, N. M. Zahr, [21], an atlas is used as prior
information in segmentation, in atlas based segmentation the image should be registered, I. Diaz and P. Boulanger[22], the proposed
method uses nesh- free total lagrangian explicit dynamic algorithm for the simulation of atlas deformation and a data driven model
of tumour using multi model MRI segmentation. The experiment performed on 12 MRIs taken from BRATS challenge training
dataset, the method introduces a novel approach of atlas to patient registration method with tumour; the method used GUP to
reduce computational time.

H. Artificial Neural Network

The ANN is based on the process of biological nervous system. it resembles the functioning of human mind, this is an information
dispensation model and are motivated by the way human intelligence system works. The basic processing element of ANN are
artificial neurons, it is adaptive as changes its configuration based on internal or external information that flows from end to end
through the system.D. M. Joshi, N. K. Rana and V. M. Misra[23], a brain cancer detection and classification system has been
designed and developed that uses computer based procedures to detect tumour and lesions and classification of tumour is done by
ANN.V. Kumar, J. Sachdeva et. al, [24] presented a method based on PCA-ANN along with GVF(Gradient vector flow) where
the detection and classification of tumours done with high accuracy rates for different cases, the overall accuracy was 91.7%.

All rights reserved by www.grdjournals.com 16

 Brain Tumour Extration From MR Images Using Segmentation Techniques: A Review
(GRDJE/ Volume 2 / Issue 4 / 003)

I. Markov Random Field

An MRF modal spatially interacts with neighbouring or nearby pixels. MRF itself is not a segmentation method but a statistical
model and can be used within segmentation method. In MRF method, the value of pixel is updated by iterated conditional modes
and simulated annealing with maximizing a posterior estimate. Usually, it is computationally expensive method. Pan Lin[25]
represented an efficient method, tissue classification method, which uses Bayesian contextual classifier based on Markov random
field models. The Bayesian theory is used to generate probability maps for each tissue, which estimates the most likely tissue
volume fraction within each voxel. All tests have been done in a ground truth image considering different noise levels. S.
Abdulbaqiet. al. [26] proposed a method on HMRF and threshold technique that segments the brain MRI and tumour is detected
by threshold technique with high accuracy.

III. COMPARARTIVE RESULTS OF DIFFERENT TUMOR DETECTION METHODS

Table 1: Literature work
Sr.
AUTHOR YEAR METHOD PRECISENESS
NO.

Successfully differentiated between affected and non-

Natarajan Prem et al.
1. 2012 Thresholding affected brain MRI.
[11]
(TUMOUR DETECTED).

Zabir, Ishmam, et al Detected tumour with accuracy between 55-95% dice

2. 2015 Region Growing
[13] coefficient.

CC Lin, CC Chang. k- means clustering with histogram Method was very efficient, tumour location detected
3. 2007
[14] statistics successfully

Benson, C. C., et al. Comparatively improved results found in detection of

4. 2016 Fuzzy c means and watershed
[16] tumour from previous one.

.K. Bhima and A.

5. 2016 Marker based Watershed Detected tumour with high accuracy.
Jagan [18]

I. Diaz and P.
6. 2015 Atlas based Tumour detected
Boulanger[22]

7. D. M. Joshi [23] 2010 Artificial neural network Tumour successfully

H. S. Abdulbaqi et. al. Hidden Markov random field with

8. 2014 Tumour segmented and detected with high accuracy.
[26] threshold techniques

IV. CONCLUSION AND FUTURE SCOPE

There are different techniques for detecting tumours like neural networks, fuzzy means, marker based, K-means and histogram,
region growing. Segmentation is faster and more accurate for tumour detection in MRI images. The segmentation is a process
which converts any gray scale image to binary image because analysis and computation on binary images are simpler. Segmentation
of medical images reduces the chances of fault and detects the tumour and lesions with good accuracy rates, different segmentation
techniques have been discussed above with some good detection rate while in each case some improvement is also required as in
future we are going to introduce a method which is more efficient to detect tumour with high accuracy and less computational
time.

REFERENCES
[1] Yan, Pingkun, and Ashraf A. Kassim. "Medical image segmentation with minimal path deformable models." Image Processing, 2004. ICIP'04. 2004
International Conference on. Vol. 4. IEEE, 2004.
[2] Ahirwar, Anamika, and R. S. Jadon. "Effectiveness of Soft Computing Techniques for Medical Imaging." Computational Intelligence and Communication
Networks (CICN), 2013 5th International Conference on. IEEE, 2013
[3] Zlch, K. J. "Principles of the new World Health Organization (WHO) classification of brain tumors." Neuroradiology 19.2 (1980): 59-66.
[4] Bushberg, Jerrold T., and John M. Boone. The essential physics of medical imaging. Lippincott Williams & Wilkins, 2011.

All rights reserved by www.grdjournals.com 17

 Brain Tumour Extration From MR Images Using Segmentation Techniques: A Review
(GRDJE/ Volume 2 / Issue 4 / 003)

[5] Roy, Vandana. "Performance Evaluation of Different Image Denoising Methods." 2012, International journal of electronics and computer science engineering
1.3 (2012): 1757-1764.
[6] Sukhatme, Neelabh, and Shailja Shukla. "Independent Component Analysis based Denoising of Magnetic Resonance Images." International Journal of
Computer Applications 54.2 (2012).
[7] Bezdek, James C., L. O. Hall, and L_P Clarke. "Review of MR image segmentation techniques using pattern recognition." Medical physics 20.4 (1993):
1033-1048.
[8] Liang, Zhengrong. "Tissue classification and segmentation of MR images." IEEE Engineering in Medicine and Biology Magazine 12.1 (1993): 81-85.
[9] Hall, Lawrence O., et al. "A comparison of neural network and fuzzy clustering techniques in segmenting magnetic resonance images of the brain." IEEE
transactions on neural networks 3.5 (1992): 672-682.
[10] Clarke, L. P., et al. "MRI segmentation: methods and applications." Magnetic resonance imaging 13.3 (1995): 343-368.
[11] Natarajan, Prem, et al. "Tumor detection using threshold operation in MRI brain images." Computational Intelligence & Computing Research (ICCIC), 2012
IEEE International Conference on. IEEE, 2012
[12] Roy, Vandana, and Shailja Shukla. "Automatic removal of artifacts from EEG signal based on spatially constrained ICA using daubechies wavelet."
International Journal of Modern Education and Computer Science 6.7 (2014): 31
[13] Zabir, Ishmam, et al. "Automatic brain tumor detection and segmentation from multi-modal MRI images based on region growing and level set evolution."
Electrical and Computer Engineering (WIECON-ECE), 2015 IEEE International WIE Conference on. IEEE, 2015.
[14] Wu, Ming-Ni, Chia-Chen Lin, and Chin-Chen Chang. "Brain tumor detection using color-based k-means clustering segmentation." Intelligent Information
Hiding and Multimedia Signal Processing, 2007. IIHMSP 2007. Third International Conference on. Vol. 2. IEEE, 2007.
[15] Chuang, Keh-Shih, et al. "Fuzzy c-means clustering with spatial information for image segmentation." computerized medical imaging and graphics 30.1
(2006): 9-15
[16] Benson, C. C., et al. "Brain tumor segmentation from MR brain images using improved fuzzy c-means clustering and watershed algorithm." Advances in
Computing, Communications and Informatics (ICACCI), 2016 International Conference on. IEEE, 2016.
[17] H. Digabel and C. Lantujoul, Iterative algorithms, in Proceedings of the 2nd European Symposium on Quantitative Analysis of Microstructures in
Materials Sciences, Biology and Medicine, pp. 8599, Caen, France, October 1977.
[18] K. Bhima and A. Jagan, "Analysis of MRI based brain tumor identification using segmentation technique," 2016 International Conference on Communication
and Signal Processing (ICCSP), Melmaruvathur, 2016, pp. 2109-2113. doi: 10.1109/ICCSP.2016.7754551
[19] I. Maiti and M. Chakraborty, "A new method for brain tumor segmentation based on watershed and edge detection algorithms in HSV colour model," 2012
NATIONAL CONFERENCE ON COMPUTING AND COMMUNICATION SYSTEMS, Durgapur, 2012, pp.1-5. doi: 10.1109/NCCCS.2012.6413020
[20] P. Dhage, M. R. Phegade and S. K. Shah, "Watershed segmentation brain tumor detection," 2015 International Conference on Pervasive Computing (ICPC),
Pune, 2015,pp.1-5. doi: 10.1109/PERVASIVE.2015.7086967
[21] T. Rohlfing, N. M. Zahr, E. V. Sullivan, & A. Pfefferbaum, The SRI24 multichannel atlas of normal adult human brain structure, Human Brain Mapp., vol.
31, no. 5, pp. 798819, May 2010.
[22] I. Diaz and P. Boulanger, "Atlas to patient registration with brain tumor based on a mesh-free method," 2015 37th Annual International Conference of the
IEEE Engineering in Medicine and Biology Society (EMBC), Milan, 2015, pp. 2924-2927. doi: 10.1109/EMBC.2015.7319004
[23] D. M. Joshi, N. K. Rana and V. M. Misra, "Classification of Brain Cancer using Artificial Neural Network," 2010 2nd International Conference on Electronic
Computer Technology, Kuala Lumpur, 2010, pp. 112-116. doi: 10.1109/ICECTECH.2010.5479975
[24] V. Kumar, J. Sachdeva, I. Gupta, N. Khandelwal and C. K. Ahuja, "Classification of brain tumors using PCA-ANN," 2011 World Congress on Information
and Communication Technologies, Mumbai, 2011, pp. 1079-1083. doi: 10.1109/WICT.2011.6141398
[25] Pan Lin, Yong Yang and Chong-Xun Zheng, "An efficient brain magnetic resonance image segmentation method," Proceedings of 2004 International
Conference on Machine Learning and Cybernetics (IEEE Cat. No.04EX826), 2004, pp. 2757-2760 vol.5. doi: 10.1109/ICMLC.2004.1378499
[26] H. S. Abdulbaqi, Mohd Zubir Mat, A. F. Omar, I. S. B. Mustafa and L. K. Abood, "Detecting brain tumor in Magnetic Resonance Images using Hidden
Markov Random Fields and Threshold techniques," 2014 IEEE Student Conference on Research and Development, Batu Ferringhi, 2014, pp. 1-5. doi:
10.1109/SCORED.2014.7072963

All rights reserved by www.grdjournals.com 18