![International Journal of Research in Advent Technology, Vol.2, No.5, May 2014
E-ISSN: 2321-9637
211
An Integrated Brain Tumour Detection Technique
Charutha S 1, M.J.Jayashree 2
P G Scholar, Department of Electronics & Communication Engineering1
Associate Professor, Department of Electronics & Communication Engineering 2,
Mar Baselios College of Engineering & Technology, Kerala University1, 2
Email: charu.sreekandan@gmail.com1, jayashreemj@yahoo.com2
Abstract- Brain tumour can be easily detected from Magnetic Resonance Imaging (MRI) images with the help
of image processing techniques. It includes several types of segmentation techniques which separates the tumour
from MRI. Here, an integrated method of brain tumour detection which combines modified texture based region
growing and edge detection is proposed. Simulation is done in MATLAB. Results show that the proposed
method is better and more accurate when compared to the individual modified texture based region growing and
edge detection. The proposed method will help to detect the tumour more efficiently.
Index Terms- Brain; Tumour detection; Segmentation; MRI; Modified texture segmentation; Edge detection.
1. INTRODUCTION
Brain tumour is a deadly disease and it can be either
benign or malignant. Different types of imaging
techniques like Magnetic Resonance Imaging (MRI),
Computed Tomography (CT) etc. are there for the
proper detection of brain tumour. From these imaging
techniques, detection of brain tumour can be done
efficiently by automatic detection. Automatic brain
tumour detection can be performed by using image
processing techniques. The most useful image
processing technique is segmentation. Different
techniques have been proposed in the area of tumour
detection using image processing. M. Usman Akram
and Anam Usman proposed global thresholding for
the brain tumour detection. Here morphological
operation is also applied after segmentation [1].
Manoj K Kowar and Sourabh Yadav proposed a novel
brain tumour detection technique which is based on
histogram thresholding. Here the threshold point of
the histograms of the two brain halves is determined
and based on that point the presence and the physical
dimension of the tumour is determined [2]. P.
Kanungo, P. K. Nanda and U. C. Samal discussed
about another segmentation technique which uses
genetic algorithm for the selection of threshold from
the histogram of MR brain images [3]. A review of
fully automated brain tumour detection techniques
from MRI images and CT images is made. Here
methods based on artificial neural network, wavelets
etc. are discussed [4]. Comparisons of different
segmentation techniques which can be used for brain
tumour detection are also done by P. K. Srimani and
Shanthi Mahesh. Here detection methods using global
thresholding, histogram clustering, watershed
segmentation and edge based segmentation were
discussed [5]. Rajesh C. Patil and Dr. A. S.
Bhalchandra proposed another brain tumour detection
technique which uses threshold segmentation and
watershed segmentation [6]. Seeded region growing
method is also implemented for the tumour detection
based on texture analysis. Texture analysis will help
to know the presence of tumour and segmented only if
the presence of tumour is detected [7]. Robert D.
Ambrosini, Peng Wang and Walter G. O’Dell
demonstrated a three-dimensional (3D) template
matching-based algorithm for detecting brain
metastases from the MR brain images [8].
In the brain tumour detection which we propose,
an integration of two types of segmentation is used.
They are modified texture based region growing
technique and classical sobel edge detection
technique. First the MRI image is segmented using
modified texture based region growing which includes
two constrains; one is the intensity constrain and other
is the texture constrain [9]. The second integrated
segmentation technique is sobel based edge detection.
We have proved that the combined segmentation
provide better results.
Organization of the paper is as follows: The
proposed technique is introduced in Section 2. The
experimental results and conclusion are given in
section 3 and section 4 respectively.
2. METHODOLOGY
The method we propose for the brain tumour
detection incorporates two segmentation techniques.
The steps in the implementation of the proposed
method are shown in Fig.1.
2.1. Image acquisition
First, MR brain images of various patients are
collected from publicly available sources. They are
further processed for detecting the tumour accurately.](https://image.slidesharecdn.com/paperid-25201482-140904043455-phpapp01/85/Paper-id-25201482-1-320.jpg)
![International Journal of Research in Advent Technology, Vol.2, No.5, May 2014
E-ISSN: 2321-9637
212
Fig.1. Flow chart of the proposed technique [9]
2.2. Pre-processing
Before segmentation, some pre-processing techniques
are applied to the MR brain images to remove the
noise present in the image [9]. The noise is removed
by high pass filtering and median filtering.
2.3. Modified texture based region growing
segmentation
Segmentation of MRI images can be done using
region growing segmentation. The normal region
growing method is a pixel based image segmentation
method. In normal region growing, only the intensity
constrain is taken into account i.e. initially a seed
point and a particular threshold level of intensity is
selected. If the difference between the intensity value
of the seed point and the neighbouring pixel is below
the selected threshold level, then those pixels are
selected for region growing. In our proposed method,
region growing technique based on two constrains is
used for segmentation. The two disadvantages of
normal region growing are over segmentation results
and difficulty in distinguishing the shading of the real
image. Since tumours have irregular shapes and
inhomogeneous structure, intensity or shape based
segmentation will be not much efficient. But they can
be segmented more accurately by their textural
properties. It increases the sensitivity of tumour
detection since intensity variation of the tumour
doesn’t affect the efficiency of tumour detection. So
the first level of segmentation stage is implemented
using modified texture based region growing. Here,
first texture filtering is done on the pre-processed
image. Based on it, a texture constrain is set in
addition to the intensity constrain present in normal
region growing. After setting the two constrains,
region growing segmentation is done [9].
2.4. Edge detection
After applying region growing segmentation, a
classical edge detection method is applied. The edge
detection used here is sobel based edge detection.
After applying sobel based edge detection, the
tumour part is extracted from the MR brain images.
Since the integration of two types of segmentation
techniques is used here, it provides better results
compared to the individual existing techniques. Also
texture based region growing provides more accurate
results compared to the normal region growing
technique.
3. RESULTS AND DISCUSSIONS
Simulation of the proposed method is done in
MATLAB. The simulated results of the proposed
method include the outputs of pre-processing stage,
modified texture segmentation stage and edge
detection stage. Also a comparison is made between
the outputs of the proposed method and manual
segmentation.
A sample MR brain image, simulated outputs of
pre-processing stage, modified texture segmentation
stage, edge detection stage and comparison of manual
segmentation and proposed method is shown in Fig.
(2-6).
Fig.2 Sample MRI image with tumour
Image Acquisition
Pre-processing
Modified texture
based region-growing
segmentation
Edge-detection
Detect Tumour](https://image.slidesharecdn.com/paperid-25201482-140904043455-phpapp01/85/Paper-id-25201482-2-320.jpg)
![International Journal of Research in Advent Technology, Vol.2, No.5, May 2014
E-ISSN: 2321-9637
213
Fig.3. Steps in pre-processing stage
Fig.4. Output of modified texture segmentation stage
Fig.5. Output of proposed method
Fig.6. Comparison of manual segmentation and output of proposed
method
From the simulated results, we can understand that
the integrated brain tumour detection technique
provides better and accurate results compared to the
manual segmentation technique.
4. CONCLUSION
Different image processing techniques are existing for
the detection of brain tumours from MRI images and
every method has its own benefits and drawbacks.
Here we propose an integrated brain tumour detection
technique which combines modified texture based
region growing segmentation and sobel based edge
detection. Experimental results have proved that it
provides accurate detection compared to the
individual classical methods and manual segmentation
methods. The future scope of the proposed technique
is that it can be extended to use it with other edge
detection methods. It can also be extended to detect
the tumours in other parts of the human body.
REFERENCES
[1] M. Usman Akram; Anam Usman. (2011):
“Computer aided system for brain tumour
detection and segmentation,” IEEE International
Conference on Computer Networks and
Information Technology, pp. 299 – 302.
[2] Manoj K Kowar; Sourabh Yadav. (2012):“Brain
tumour detection and segmentation using
histogram thresholding,” International Journal of
Engineering and Advanced Technology, vol.1,
pp.16-20.
[3] P. Kanungo; P. K. Nanda; U. C. Samal: “Image
segmentation using thresholding and genetic
algorithm.”
[4] Anjum Hayat Gondal; Muhammad Naeem
Ahmed Khan. (2013): “A review of fully
automated techniques for brain tumour detection
from MR images,” International Journal of
Modern Education and Computer Science, vol.5,
pp. 55-61.](https://image.slidesharecdn.com/paperid-25201482-140904043455-phpapp01/85/Paper-id-25201482-3-320.jpg)
![International Journal of Research in Advent Technology, Vol.2, No.5, May 2014
E-ISSN: 2321-9637
214
[5] P.K.Srimani; Shanthi Mahesh. (2013): “A
comparative study of different segmentation
techniques for brain tumour detection”,
International Journal of Emerging Technologies
in Computational and Applied Sciences, vol.4,
pp. 192-197.
[6] Rajesh C. Patil; Dr. A. S. Bhalchandra: “Brain
tumour extraction from MRI images using
MATLAB”, International Journal of Electronics,
Communication & Soft Computing Science and
Engineering, vol.2, pp. 1-4.
[7] Mukesh Kumar; Kamal K.Mehta. (2011): “A
texture based tumour detection and automatic
segmentation using seeded region growing
method”, International Journal of Computer
Technology and Applications, vol.2, pp.855-859.
[8] Robert D. Ambrosini; Peng Wang; Walter G.
O’Dell. (2010): “Computer-aided detection of
metastatic brain tumours using automated three-dimensional
template matching”, Journal of
Magnetic Resonance Imaging, pp.85-93.
[9] K. S. Angel Viji; Dr J. Jayakumari. (2013):
“Modified texture based region growing
segmentation of MR brain images”, IEEE
Conference on Information and Communication
Technologies, pp.691-695.](https://image.slidesharecdn.com/paperid-25201482-140904043455-phpapp01/85/Paper-id-25201482-4-320.jpg)
Paper id 25201482
- 1. International Journal of Research in Advent Technology, Vol.2, No.5, May 2014 E-ISSN: 2321-9637 211 An Integrated Brain Tumour Detection Technique Charutha S 1, M.J.Jayashree 2 P G Scholar, Department of Electronics & Communication Engineering1 Associate Professor, Department of Electronics & Communication Engineering 2, Mar Baselios College of Engineering & Technology, Kerala University1, 2 Email: [email protected], [email protected] Abstract- Brain tumour can be easily detected from Magnetic Resonance Imaging (MRI) images with the help of image processing techniques. It includes several types of segmentation techniques which separates the tumour from MRI. Here, an integrated method of brain tumour detection which combines modified texture based region growing and edge detection is proposed. Simulation is done in MATLAB. Results show that the proposed method is better and more accurate when compared to the individual modified texture based region growing and edge detection. The proposed method will help to detect the tumour more efficiently. Index Terms- Brain; Tumour detection; Segmentation; MRI; Modified texture segmentation; Edge detection. 1. INTRODUCTION Brain tumour is a deadly disease and it can be either benign or malignant. Different types of imaging techniques like Magnetic Resonance Imaging (MRI), Computed Tomography (CT) etc. are there for the proper detection of brain tumour. From these imaging techniques, detection of brain tumour can be done efficiently by automatic detection. Automatic brain tumour detection can be performed by using image processing techniques. The most useful image processing technique is segmentation. Different techniques have been proposed in the area of tumour detection using image processing. M. Usman Akram and Anam Usman proposed global thresholding for the brain tumour detection. Here morphological operation is also applied after segmentation [1]. Manoj K Kowar and Sourabh Yadav proposed a novel brain tumour detection technique which is based on histogram thresholding. Here the threshold point of the histograms of the two brain halves is determined and based on that point the presence and the physical dimension of the tumour is determined [2]. P. Kanungo, P. K. Nanda and U. C. Samal discussed about another segmentation technique which uses genetic algorithm for the selection of threshold from the histogram of MR brain images [3]. A review of fully automated brain tumour detection techniques from MRI images and CT images is made. Here methods based on artificial neural network, wavelets etc. are discussed [4]. Comparisons of different segmentation techniques which can be used for brain tumour detection are also done by P. K. Srimani and Shanthi Mahesh. Here detection methods using global thresholding, histogram clustering, watershed segmentation and edge based segmentation were discussed [5]. Rajesh C. Patil and Dr. A. S. Bhalchandra proposed another brain tumour detection technique which uses threshold segmentation and watershed segmentation [6]. Seeded region growing method is also implemented for the tumour detection based on texture analysis. Texture analysis will help to know the presence of tumour and segmented only if the presence of tumour is detected [7]. Robert D. Ambrosini, Peng Wang and Walter G. O’Dell demonstrated a three-dimensional (3D) template matching-based algorithm for detecting brain metastases from the MR brain images [8]. In the brain tumour detection which we propose, an integration of two types of segmentation is used. They are modified texture based region growing technique and classical sobel edge detection technique. First the MRI image is segmented using modified texture based region growing which includes two constrains; one is the intensity constrain and other is the texture constrain [9]. The second integrated segmentation technique is sobel based edge detection. We have proved that the combined segmentation provide better results. Organization of the paper is as follows: The proposed technique is introduced in Section 2. The experimental results and conclusion are given in section 3 and section 4 respectively. 2. METHODOLOGY The method we propose for the brain tumour detection incorporates two segmentation techniques. The steps in the implementation of the proposed method are shown in Fig.1. 2.1. Image acquisition First, MR brain images of various patients are collected from publicly available sources. They are further processed for detecting the tumour accurately.
- 2. International Journal of Research in Advent Technology, Vol.2, No.5, May 2014 E-ISSN: 2321-9637 212 Fig.1. Flow chart of the proposed technique [9] 2.2. Pre-processing Before segmentation, some pre-processing techniques are applied to the MR brain images to remove the noise present in the image [9]. The noise is removed by high pass filtering and median filtering. 2.3. Modified texture based region growing segmentation Segmentation of MRI images can be done using region growing segmentation. The normal region growing method is a pixel based image segmentation method. In normal region growing, only the intensity constrain is taken into account i.e. initially a seed point and a particular threshold level of intensity is selected. If the difference between the intensity value of the seed point and the neighbouring pixel is below the selected threshold level, then those pixels are selected for region growing. In our proposed method, region growing technique based on two constrains is used for segmentation. The two disadvantages of normal region growing are over segmentation results and difficulty in distinguishing the shading of the real image. Since tumours have irregular shapes and inhomogeneous structure, intensity or shape based segmentation will be not much efficient. But they can be segmented more accurately by their textural properties. It increases the sensitivity of tumour detection since intensity variation of the tumour doesn’t affect the efficiency of tumour detection. So the first level of segmentation stage is implemented using modified texture based region growing. Here, first texture filtering is done on the pre-processed image. Based on it, a texture constrain is set in addition to the intensity constrain present in normal region growing. After setting the two constrains, region growing segmentation is done [9]. 2.4. Edge detection After applying region growing segmentation, a classical edge detection method is applied. The edge detection used here is sobel based edge detection. After applying sobel based edge detection, the tumour part is extracted from the MR brain images. Since the integration of two types of segmentation techniques is used here, it provides better results compared to the individual existing techniques. Also texture based region growing provides more accurate results compared to the normal region growing technique. 3. RESULTS AND DISCUSSIONS Simulation of the proposed method is done in MATLAB. The simulated results of the proposed method include the outputs of pre-processing stage, modified texture segmentation stage and edge detection stage. Also a comparison is made between the outputs of the proposed method and manual segmentation. A sample MR brain image, simulated outputs of pre-processing stage, modified texture segmentation stage, edge detection stage and comparison of manual segmentation and proposed method is shown in Fig. (2-6). Fig.2 Sample MRI image with tumour Image Acquisition Pre-processing Modified texture based region-growing segmentation Edge-detection Detect Tumour
- 3. International Journal of Research in Advent Technology, Vol.2, No.5, May 2014 E-ISSN: 2321-9637 213 Fig.3. Steps in pre-processing stage Fig.4. Output of modified texture segmentation stage Fig.5. Output of proposed method Fig.6. Comparison of manual segmentation and output of proposed method From the simulated results, we can understand that the integrated brain tumour detection technique provides better and accurate results compared to the manual segmentation technique. 4. CONCLUSION Different image processing techniques are existing for the detection of brain tumours from MRI images and every method has its own benefits and drawbacks. Here we propose an integrated brain tumour detection technique which combines modified texture based region growing segmentation and sobel based edge detection. Experimental results have proved that it provides accurate detection compared to the individual classical methods and manual segmentation methods. The future scope of the proposed technique is that it can be extended to use it with other edge detection methods. It can also be extended to detect the tumours in other parts of the human body. REFERENCES [1] M. Usman Akram; Anam Usman. (2011): “Computer aided system for brain tumour detection and segmentation,” IEEE International Conference on Computer Networks and Information Technology, pp. 299 – 302. [2] Manoj K Kowar; Sourabh Yadav. (2012):“Brain tumour detection and segmentation using histogram thresholding,” International Journal of Engineering and Advanced Technology, vol.1, pp.16-20. [3] P. Kanungo; P. K. Nanda; U. C. Samal: “Image segmentation using thresholding and genetic algorithm.” [4] Anjum Hayat Gondal; Muhammad Naeem Ahmed Khan. (2013): “A review of fully automated techniques for brain tumour detection from MR images,” International Journal of Modern Education and Computer Science, vol.5, pp. 55-61.
- 4. International Journal of Research in Advent Technology, Vol.2, No.5, May 2014 E-ISSN: 2321-9637 214 [5] P.K.Srimani; Shanthi Mahesh. (2013): “A comparative study of different segmentation techniques for brain tumour detection”, International Journal of Emerging Technologies in Computational and Applied Sciences, vol.4, pp. 192-197. [6] Rajesh C. Patil; Dr. A. S. Bhalchandra: “Brain tumour extraction from MRI images using MATLAB”, International Journal of Electronics, Communication & Soft Computing Science and Engineering, vol.2, pp. 1-4. [7] Mukesh Kumar; Kamal K.Mehta. (2011): “A texture based tumour detection and automatic segmentation using seeded region growing method”, International Journal of Computer Technology and Applications, vol.2, pp.855-859. [8] Robert D. Ambrosini; Peng Wang; Walter G. O’Dell. (2010): “Computer-aided detection of metastatic brain tumours using automated three-dimensional template matching”, Journal of Magnetic Resonance Imaging, pp.85-93. [9] K. S. Angel Viji; Dr J. Jayakumari. (2013): “Modified texture based region growing segmentation of MR brain images”, IEEE Conference on Information and Communication Technologies, pp.691-695.