Brain Tumor Extraction Using Genetic Algorithm

International Journal on Future Revolution in Computer Science & Communication Engineering ISSN: 2454-4248
Volume: 4 Issue: 6 33 – 39
_______________________________________________________________________________________________
Brain Tumor Extraction using Genetic Algorithm
Arti Gujar, Prof. C. M. Meshram

Department of CE
Bapurao Deshmukh College of Engineering,
Sevagram, Wardha, Maharashtra, India
Abstract: A Brain Cancer is very serious disease causing deaths of many individuals. The detection and classification system must be available
so that it can be diagnosed at early stages. Cancer classification has been one of the most challenging tasks in clinical diagnosis. At present
cancer classification is done mainly by looking through the cells’ morphological differences, which do not always give a clear distinction of
cancer subtypes. Unfortunately, this may have a significant impact on the final outcome of whether a patient could be cured effectively or not.
We have proposed a methodology to segment and classify the brain MRI image using k-means clustering algorithm and Genetic algorithm.
Keywords: k-means, Genetic classification
__________________________________________________*****_________________________________________________
Introduction Segmentation of brain tissues in gray matter, white

Automated classification and detection of matter and tumor on medical images is not only of high
tumors indifferent medical images is motivated by the interest in serial treatment monitoring of “disease burden” in
necessity of high accuracy when dealing with a human oncologic imaging, but also gaining popularity with the
life. Also, the computer assistance is demanded in advance of image guided surgical approaches. Outlining the
medical institutions due to the fact that it could improve brain tumor contour is a major step in planning spatially
the results of humans in such a domain where the false localized radiotherapy (e.g., Cyber knife, iMRT ) which is
negative cases must be at a very low rate. It has been usually done manually on contrast enhanced T1-weighted
proven that double reading of medical images could lead magnetic resonance images (MRI) in current clinical
to better tumor detection. Conventional methods of practice. On T1 MR Images acquired after administration of
monitoring and diagnosing the diseases rely on detecting a contrast agent (gadolinium), blood vessels and parts of the
the presence of particular features by a human observer. tumor, where the contrast can pass the blood–brain barrier
Due to large number of patients in intensive care units are observed as hyper intense areas. There are various
and the need for continuous observation of such attempts for brain tumor segmentation in the literature
conditions, several techniques for automated diagnostic which use a single modality, combine multi modalities and
systems have been developed in recent years to attempt use priors obtained from population atlases.
to solve this problem. Such techniques work by
transforming the mostly qualitative diagnostic criteria Literature Review
into a more objective quantitative feature classification Parveen, Amritpal singh suggested that method uses Gray
problem. In this project the automated classification of Level Run Length Matrix (GLRLM) to extract features.The
brain magnetic resonance images by using some prior reduced GLRLM features are deferred to support vector
knowledge like pixel intensity and some anatomical machine for training and testing.The brain MRI images were
features is proposed. Currently there are no methods classified using SVM techniques which are widely used for
widely accepted therefore automatic and reliable data analyzing and pattern recognizing[1].
methods for tumor detection are of great need and Arashdeep Kaur suggested that method used an approach to
interest. The application of PNN in the classification of detect brain tumor using four different methods namely
data for MR images problems are not fully utilized yet. Otsu, K-means, Fuzzy-c-Means and thresholding. The main
These included the clustering and classification objective of this paper is to develop a fully automated brain
techniques especially for MR images problems with huge tumor detection system that can detect and extract tumor
scale of data and consuming times and energy if done from MR Image of brain. This paper also gives the
manually. Thus, fully understanding the recognition, comparison between the algorithms presented [2].
classification or clustering techniques is essential to the Aniket A. kathalkar suggested that deals with such a
developments of Neural Network systems particularly in system which uses computer based procedures to detect
medicine problems. tumor blocks and classify the type of tumor using Artificial
33
IJFRCSCE | June 2018, Available @ http://www.ijfrcsce.org
_______________________________________________________________________________________
_______________________________________________________________________________________________
Neural Network Algorithm for MRI images of different Eltaher Mohamed Hussein 1 suggested an automated
patients. Different image processing techniques such as recognition system for MR imaging using ANNs. It was
histogram equalization, image segmentation, image observed that when Elman network was used during the
enhancement, morphological operations and recognition process, the duration time and the accuracy level
featureextraction are used for detection of the brain tumor in were high, compared with other ANNs systems.[7]
the MRI images of the cancer affected patients[3] Kamal Kant Hiran reviewed the techniques of the MRI
Aqhsa Q. Syed1,K.Narayanan introduced one automatic image enhancement in terms of tumor pixels detected. They
brain tumor detection method to increase the accuracy and studied several digital image processing methods and
yield and decrease the diagnosis time. Here, it is tried to discussed its requirements and properties in brain tumor
give clear description from brain tissues using Multi-Layer detection .This paper gives enhanced information about
Perception Network, energy, entropy, contrast and some brain tumor detection and segmentation. The marked area is
other statistic features such as mean, median, variance and segmented and the assessment of this tool from the
correlation. It is used from a feature selection method to radiologist, whom the project is concerned with, is positive
reduce the feature space too. This method uses from neural and this tool helps them in diagnosis, the treatment
network to do this classification.[4] procedure and state of the tumor monitoring.[8]
S. N. Deepa and B. Aruna Devi introduced a survey has R. J. Deshmukh, R.S. Khule presents a automated
been made on the applications of intelligent computing recognition system for the MRI image using the neuro fuzzy
techniques for diagnostic sciences in biomedical image logic. Features are extracted from raw images which are
classification. This study gathers representative works that then fed to ANFIS (Artificial neural fuzzy inference
exhibit how AI is applied to the solution of very different system).ANFIS being neuro-fuzzy system harness power of
problems related to different diagnostic science analysis. It both hence it proves to be a sophisticated framework for
also detects the methods of artificial intelligence that are multi object classification. A comprehensive feature set and
used frequently together to solve the special problems of fuzzy rules are selected to classify an abnormal image to the
medicine. SVM neural network issued in almost all imaging corresponding tumor type.[9]
modalities of medical image classification. Similarly fuzzy The system develoved in this study classifies and identifies
C means and improvements to it are important tool in pathological tissues in a noninvasive and automated fashion.
segmentation of brain images. Various diagnostic studies The designed brain tumor detection and classification
like mammogram analysis, MRI brain analysis, bone and system uses conceptually simple classification method using
retinal analysis etc., using neural network approach result in the neural network. Textures features are used in the training
use of back propagation network, probabilistic neural of the ANN. Co-occurrence matrices at different directions
network, and extreme learning machine recurrently. Hybrid are calculated and GLCM features and Gabor features are
approach of GA and PSO are also commonly used for extracted from the matrices.[10]
feature extraction and feature selection [5]
J. selvakumar, A. Lakshmi,T.Arivoli suggested that a Proposed System Architecture
noise free image is given as a input to the k-means and The method include brain tumor segmentation using
tumour is extracted from the MRI image. And then optimized k-means and genetic algorithm.
segmentation using Fuzzy C means for accurate tumour Optimized k-means algorithm will be used for clustering.
shape extraction of malignant tumour and thresholding of Genetic algorithm will be used for classification. At the end
output in feature extraction. Finally approximate reasoning of process the tumor is extracted from the MRI image and
for calculating tumour shape and position calculation. The also its exact position and shape are determined in grayscale
experimental results are compared with other algorithms.[6]
34
_______________________________________________________________________________________
_______________________________________________________________________________________________
MRI Brain Fast discrete Texture features Trained Network

Image curvelet Extraction
Decomposition
Training Feature Extraction &

Samples PNN-RBF Training
Classification
Performance Tumor and structural Segmentation If Input Is

analysis part detection Process Abnormal
Fig: System Architecture
For optimum classification purposes, what concern us are that satisfy P. C[i][j] is a measure of the joint
the statistical techniques of characterization… This is probability that a pair of points satisfying P
because it is these techniques that result in computing will have values g[i], g[j].
texture properties… The most popular statistical - C is called a co-occurrence matrix defined by
representations of texture are: P.
(1) Co-occurrence Matrix Examples for the operator P are: “i above j”, or “i one
(2) Tamura Texture position to the right and two below j”, etc.
This can also be illustrated as follows… Let t be a
(3) Wavelet Transform translation, then a co-occurrence matrix Ctof a region is
defined for every grey-level (a, b) by [1]:
Co-Occurrence Matrix
Ct (a,b) card{(s,s t) R2| A[s] a, A[s t]
Originally proposed by R.M. Haralick, the co-occurrence
b}
matrix representation of texture features explores the grey
Here, Ct(a, b) is the number of site-couples, denoted by (s, s
level spatial dependence of texture. A mathematical
+ t) that are separated by a translation vector t, with a being
definition of the co-occurrence matrix is as follows:
the grey-level of s, and b being the grey-level of s + t.
- Given a position operator P(i,j), For example; with an 8 grey-level image representation
- let Abe an n x n matrix and a vector t that considers only one neighbor, we
would find [1]:
- Whose element A[i][j] is the number of times
that points with grey level (intensity) g[i]
occur, in the position specified by P, relative to
points with grey level g[j].
- Let C be the n x n matrix that is produced by
dividing A with the total number of point pairs
35
_______________________________________________________________________________________
_______________________________________________________________________________________________
Classical Co-Occurrence Matrix parameters.Above image shows four accuracy that can be
At first the co-occurrence matrix is constructed, based on calculated .
the orientation and distance between image pixels. Then Result Analysis
meaningful statistics are extracted from the matrix as the Performance Measurement
texture representation. Haralick proposed the following The performance of the given system measured by using the
texture features: following table that given various condition for outputs.
(1) Energy They give the multiple images of brain MRI that are
detected by the system. The first condition if we provide one
(2) Contrast brain MRI image from the dataset and system detected it as
(3) Correlation tumor is present then we consider the true positive. If we
(4) Homogeneity provide another image from the dataset and system detected
Hence, for each Haralick texture feature, we obtain a is tumor is absent then we consider the false negative. For
co-occurrence matrix. These co-occurrence matrices calculating a precision, we required a relevant dataset and
represent the spatial distribution and the dependence of for the recall we take the overall dataset of the given system.
the grey levels within a local area. Each (i,j) th entry in the For the result analysis we provide the number of tumor
matrices, represents the probability of going from one present and tumor absent and find the precision and recall
pixel with a grey level of 'i' to another with a grey level of by the following formula.
'j' under a predefined distance and angle. From these Precision = Relevant instance / Retrieve instance
matrices, sets of statistical measures are computed, called Recall = Relevant instance that have been retrieve / Total
feature vectors. amount of instance
Figure: Final Output of the System

Above image shows the tumor segmentation and
classification. It shows type of tumor and its
Function Accuracy Precision Recall

Linear 91.66% 100% 83.33%
Quadractic 83.33% 100% 66.66%
Polygonal 87.50% 100% 75.00%
Table : Performance Evolution for Existing System
Function Accuracy Precision Recall

Linear 93% 100% 93%
Quadractic 87% 100% 91%
Polygonal 89% 100% 92.5%
Table : Performance Evolution for proposed System
36
_______________________________________________________________________________________
_______________________________________________________________________________________________
Evaluation Result system according to the accuracy from the 23 images which
The graph shows 6.1 the result of precision and recall of the are taken from the dataset.
brain MRI image of the current system as well as existing
120.00%
100.00%
80.00%
Accuracy
60.00%
Precision
Recall
40.00%
20.00%
0.00%
Existing Proposed
Fig: Graph based on Linear accuracy
120.00%
100.00%
80.00%
Accuracy
60.00%
Precision
Recall
40.00%
20.00%
0.00%
Existing Proposed
Fig : Graph based on Quadractic Accuracy
37
_______________________________________________________________________________________
_______________________________________________________________________________________________
120.00%
100.00%
80.00%
Accuracy
60.00%
Precision
Recall
40.00%
20.00%
0.00%
Existing Proposed
Fig : Graph based on Polygonal Accuracy
Table: Percentage of Tumor Classification
Sr.No. Tumor Type Precision Recall

1 Benign Tumor 100% 47.82%
2 Malignant 100% 52.17%
Tumor
120%
100%
80%
60% Precision
Recall
40%
20%
0%
Benign Tumor Malignant Tumor
Fig: Graph for Tumor Classifications
38
_______________________________________________________________________________________
_______________________________________________________________________________________________
Conclusion [8]. R. J.Deshmukh, R.S Khule," Brain Tumour Detection
The project presented that automated brain Using Artificial Neural Network Fuzzy Inference System
image classification for early stage abnormality detection (ANFIS)", International Journal of Computer Applications
Technology and Research Volume 3- Issue 3, 150- 154,
with use of neural network classifier and spotting of
2014, ISSN: 2319-8656.
tumor was done with image segmentation. Pattern
[9]. Khateeja Ambareen, M. S. Mallikarjuna Swamy, Dr.
recognition was performed using probabilistic neural Rajesh Raman, "Astrocytoma Type of Brain Tumour
network with radial basis function and pattern will be Classification using Artificial Neural Network",
characterized with the help of fast discrete curvelet International Journal of Electronics Communication and
transform and haralick features analysis. From an Computer Engineering ,Volume 5, Issue 2, ISSN (Online):
experiment, system proved that it provides better 2249-071X, ISSN (Print): 2278-4209.
classification accuracy with various stages of test [10]. Monica Subashini.M #1, Sarat Kumar Sahoo, "Brain MR
samples and it consumed less time for process. Image Segmentation for Tumour Detection using Artificial
Neural Networks", Vol 5 No 2 Apr-May 2013, ISSN :
0975-4024.
Future Work
[11]. Ms. Sangeetha C., Ms. Shahin A., "BRAIN TUMOUR
Currently our proposed system works on only two types of SEGMENTATION USING ARTIFICIAL NEURAL
brain tumor. In future ,the proposed system can be improved NETWORK", International Research Journal of
by adding different different type tumor for classification. Engineering and Technology (IRJET), Volume: 02 Issue:
Also the accuracy of the proposed system can be improved 04 | July2015.
by including different different dataset available from
doctors throughtout the world.The proposed system can also
be further extended for detecting cancer and other tumors.
References
[1]. Arashdeep Kaur ,An Automatic Brain Tumour Extraction
System using Different Segmentation Methods", 2016
Second International Conference on Computational
Intelligence & Communication Technology.
[2]. Aniket A. Kathalkar, R. S. Kawitkar, Amruta Chopade ,"
Artificial Neural Network based Brain Cancer Analysis and
Classification", International Journal of Computer
Applications (0975-8887) Volume 66- No.10, March 2013.
[3]. Aqhsa Q. Syed1, K. Narayanan ," Detection of Tumour in
MRI Images Using Artificial Neural Networks",
International Journal of Advanced Research in Electrical,
Electronics and Instrumentation Engineering, Vol. 3, Issue
9, September 2014.
[4]. S.N. Deepa and B. Aruna Devi," A survey on artificial
intelligence approaches for medical image classification",
Indian Journal of Science and Technology, Vol. 4 No. 11
(Nov 2011) ISSN: 0974- 6846
[5]. J. selvakumar, A. Lakshmi, T. Arivoli," Brain Tumour
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 (ICAESM-
2012) March 30, 31, 2012.
[6]. Eltaher Mohamed Hussein1, Dalia Mahmoud Adam
Mahmoud2," Brain Tumour Detection Using Artificial
Neural Networks", Journal of Science and Technology Vol.
13, No. 2 ISSN 1605-427X Engineering and Computer
Sciences (ECS).
[7]. Kamal Kant Hiran1, Ruchi Doshi," An Artificial Neural
Network Approach for Brain Tumour Detection Using
Digital Image Segmentation", International Journal
ofEmerging Trends & Technology in Computer Science
(IJETTCS), Volume 2, Issue 5, September-October 2013.
39
_______________________________________________________________________________________

Brain Tumor Extraction Using Genetic Algorithm

Uploaded by

Brain Tumor Extraction Using Genetic Algorithm

Uploaded by

International Journal on Future Revolution in Computer Science & Communication Engineering ISSN: 2454-4248

Brain Tumor Extraction using Genetic Algorithm

Arti Gujar, Prof. C. M. Meshram

Keywords: k-means, Genetic classification

Introduction Segmentation of brain tissues in gray matter, white

MRI Brain Fast discrete Texture features Trained Network

Training Feature Extraction &

Performance Tumor and structural Segmentation If Input Is

Fig: System Architecture

Figure: Final Output of the System

Function Accuracy Precision Recall

Function Accuracy Precision Recall

Fig: Graph based on Linear accuracy

Fig : Graph based on Quadractic Accuracy

Fig : Graph based on Polygonal Accuracy

Table: Percentage of Tumor Classification

Sr.No. Tumor Type Precision Recall

Fig: Graph for Tumor Classifications

You might also like