Brain Tumor Detection using MRI: A Review of

advertisement
Brain Tumor Detection using MRI: A Review of Literature
Priyanka S. Jadhav
Meeta Bakuli
Assistant Professor, Dept. of E&TC
Engg, G.H. Raisoni College of Engg.
and Management, University of Pune,
Pune, India
PG Student, Dept.of E&TC Engg.,
G.H. Raisoni College of Engg. and
Management, University of Pune,
Pune, India
Meeta.bakuli@raisoni.net
jpriyanka211191@gmail.com
ABSTRACT
Currently, multimodal MRI images are mostly used for
segmenting brain tumor image, because MRI image gives
more information on tumor and its regions. But modern
medical imaging research faces difficulty in detection of
tumor through MRI. In detection of brain tumor using MRI,
the segmentation of brain tumor is most important task. For
this purpose brain tumor image is divided into number of
regions. This is a important procedure but very difficult in
detecting tumor. For best results it is necessary that the
segmentation must be done correctly and accurately. There are
various algorithms available for detection and segmentation of
brain tumor. However this paper gives quick review of
techniques which are developed for detection of brain tumor
using MRI. At last conclusion of this paper provides direction
towards advanced research studies on brain tumor detection.
Keywords
Brain Tumor, Magnetic Resonance Image (MRI),
Preprocessing and Enhancement, Segmentation, Feature
Extraction, Classification.
1. INTRODUCTION
A brain tumor is nothing but a intracranial neoplasm. It occurs
when abnormal cells form in the brain. According to IARC
(International Agency for Research on Cancer), nearly 126000
people faces brain tumor problem per year around the world
[1]. To overcome this problem more peoples are doing
research on the brain tumor detection and segmentation. There
are two most common tests used for the finding presence of
brain tumor. First one is MRI (Magnetic Resonance Imaging)
and CT (Computer tomography). There are many treatment
options available which include radiation therapy surgery,
chemotherapy. The selection of treatment option is based on
size, grade and type of brain tumor. For improving accuracy
many people go for CAD. Actually the output of computer is
not a first opinion, it gives output to assist radiologist to
minimize image reading time. The segmentation of brain
tumor is difficult task because tumor has variety of shapes and
sizes [2]. Another factor is image intensity, because tumor
present at different location has different image intensities [3].
These factors make segmentation process difficult. This paper
presents a review of techniques used for brain tumor detection
using MRI. The paper gives conclusion with the discussion on
future trends of advance research on brain segmentation.
2. OVERVIEW OF BRAIN TUMOR AND
MRI
2.1 Brain Tumor
All Brain tumor is nothing but abnormal mass of brain tissue
which interrupts the normal functioning of brain and creates
increasing pressure [2]. The brain tumor is classified
according to the type, size and location. There are 120 types
of brain tumor classified by World Health Organization
(WHO). This classification is based on behaviour of the cell.
There are two main types of tumor primary and secondary
brain tumor. Primary tumor is originated in brain and
secondary tumors originate from another part of body.
2.2 MR Imaging
There are number of imaging techniques which are useful for
the study of brain tumor such as Computed Tomography
(CT), MRI (Magnetic Resonance Imaging), Single Photon
Emission Computer Tomography (SPECT), Positron
Emission Tomography (PET). Now a days CT and MRI
techniques are most popular techniques. The contrast
resolution of MRI is higher than other techniques. MRI
devices can generate 3D space images. For MR imaging,
patient need to placed in strong magnetic field which cause
the proton align in parallel or antiparallel orientation with
magnetic field. After this the spinning proton are move out of
equilibrium state by introducing radio frequency pulse. The
proton returns to equilibrium state when we remove or stop
radio frequency pulse also it produces sinusoidal signal. This
generated signal is detected by scanner and final image will be
created [4].
Fig 1: MRI Scanner Cutaway
3. LITERATURE REVIEW
3.1 Image Segmentation
Image segmentation is most difficult task. Differentiating the
normal and abnormal area is most challenging task while
segmenting image. Image segmentation is classified into three
types:
1
3.1.1 Spatial Clustering
There is a difference between image segmentation and image
clustering. In image segmentation grouping is done in spatial
domain and In image clustering it is done in measurement
space.
3.1.2 Split and Merge Segmentation
4.1.3 Marker Controlled Watershed [18]

Different values of threshold are selected for
creating the Marker Controlled Watershed.

Threshold values are highly dependent on shape and
size of tumor and also on the view points (axial,
coronal) of images.
In split method, firstly entire image is consider and it splits
into quarters. This process is repeated until a homogeneity
criterion is not satisfied. In merge method, it merge or joins
adjacent segments of the same object [5].
3.1.3 Region Growing
In region growing method neighbouring points are connected
to each other to make the region bigger. This method is
mostly depend on the selection of threshold value [5,6,7].
4. REVIEW OF TECHNIQUES
AVAILABLE FOR IMAGE
SEGMENTATION
The manual brain tumor segmentation is time consuming
process to overcome this problem all people mostly choose
automated detection and segmentation. In recent years, many
methods are developed for automation of imaging, scanning,
detection and segmentation. These methods mainly classified
into intelligent based method and non-intelligent based
method. Out of which some methods have been reviewed.
4.1 Thresholding based method
It is a non-intelligent based method. In this method firstly
intensity value is calculated which is called threshold which
separate each class from other. This method is mainly
depending on gray level intensity value. In this method the
pixel having intensity greater than threshold are group into
one class and remaining into other class. But the drawback of
this method is there are only two classes. Also this method is
not good when borders are not clear [8]. Thresholding based
methods are as followed:
4.1.1 Threshold and outlier detection and
Gaussian models [16]

The edema region may require secondary analysis
followed by treatment after the primary focus on the
tumor region.

The detecting technique uses a concept to detect
difference between normal and abnormal space.

Intensity features are used in this technique.
4.1.2 Probability level set evolution [17]

The automatic method has a lower level of
agreement with the human experts compared to the
semi-automatic method.

Only two MR images samples are used for testing
and evaluate.
Fig.2 Result of Thresholding for brain MRI
4.2 Region Growing based method
This technique is used to extract a region of the image which
is based on predefined criteria. The manually selected seed
points are required for region growing and extract all pixels
which are connected to initial seed. The drawback of this
technique is it is sensitive to noise so this method is not
sufficient for segmenting brain tumor accurately. This method
is semiautomatic method because it requires manual input
seed selection. Segmentation based on region growing is
simpler than edge detection method. There are two methods of
segmentation based on region:
4.2.1 Region Growing
Steps for region growing processing:
1)Select group of seed pixels from original image [9].
2)Select set of identical criterion and develop stopping rule.
3)For grow regions attach each seed neighbouring pixels who
are having similar properties to seed pixel.
4)If no more pixels meet the criterion then control the
grouping region.
4.2.2 Region Splitting and Merging
In region growing we select seed point but in this method user
may divide image into set of detached regions and merge that
regions [10, 11] to satisfy the conditions. This method is
based on quad tree data.
2
4.4 Fuzzy based method
Fig.3. Brain Tumor Detection Using Region Growing based
method
Fuzzy logic is based on degree of membership rather than
crisp membership. This algorithm helps to perform tasks
related to intelligent tumor behaviours. The membership
function is obtained by using intensity histogram analysis and
knowledge extraction. This method gives lowest score of 71%
and highest score 93%. Dunn uses fuzzy c-mean (FCM)
clustering algorithm for image segmentation [12].FCM is
implemented by several researchers. In clustering most
important task is finding cluster to classify pixels [13]. FCM
algorithm is fast, unsupervised and simpler than other
algorithm. Jaffar introduce a FCM using curvelet transform to
remove the noise [14].
4.4.1 Fuzzy models Image Fusion [22]
4.3 Neural Network based method
This method uses a model which consist neurons and weighed
connection between them. To obtain the values of weights i.e.
coefficients requires training. Several types of network used
for image segmentation for eg. Multilayer Perceptron (MLP).
The earliest application of MLP are trained the neural network
using known diagnostic image. Detection and visualization of
brain tumor can be done by using unsupervised learnining
artificial neural network. The hierarchical Self Organizing
Map (HSOM) is a modified version of SOM.
4.3.1 ANN using canny edge detection adaptive
thresholding [20]

MRI brain represents a healthy brain or a tumor
brain.

In this paper two types of input features are used:
canny edge detection and adaptive thresholding
4.3.2 Probability Discrete wavelet transforms
(DWT) back propagation (BPNN) [21]

This paper uses hybrid classifier to distinguish
normal and abnormal brain

This method had good accuracy in classifying both
training and testing images.

This method may be used for MRI images with
other contrast mechanisms like TI-weighted, protondensity weighted and diffusion-weighted MRIs.

The proposed algorithm consists of: the registration
of multispectral MR images, the creation of fuzzy
models describing the characteristics of tumor, the
fusion based on fuzzy fusion operators and the
adjustment by fuzzy region growing based on fuzzy
connecting.

Using linear image registration tool for evaluate the
proposed method.
4.4.2 Histogram- based Fuzzy c mean smoothen
the boundaries [23]

The FSL library tool based software was compared
with the performance of the proposed algorithm. It
is considered to be a good candidate for fully
automatic MRI analysis systems.

Because of the fully automated nature of the
algorithm with no human intervention, along with
lesser number of iterations taken
4.4.3 Fuzzy C Means [24]

Generalized spatial fuzzy c-means (CSFCM)
algorithm which possesses both pixel attributes and
spatial local information that is weighted in
correspondence with neighbor elements based on
their distance attributes.

This has the potentiality to improve
segmentation performance tremendously.

This improves the segmentation performance
dramatically.

Poor contrast, noise and non-uniform intensity
variation can affect the results.
the
Fig.4. Brain Tumor Detection Using Neural Network
3

The thresholds taken from the EM output are
utilized to adaptively adjusting the rate of expansion
in the segmented area. This increases the
classification quality.
Fig.6. Brain Tumor Detection Using Hybrid based method
Fig.5. Brain Tumor Detection Using Fuzzy based method
4.5 Hybrid based technique
Different methods of machine learning algorithm are
combined to form hybrid system. It gives better solution to a
problem, which is not given by single method. Jzau-sheng
uses global histogram of images and constructed neurons.
Nandita proposed a method in which input is fuzzy and output
is crisp value [15]. But the intensity values are not fix so that
there is a difficulty for ascertaining generalized efficiency.
4.5.1 ANN fuzzy logic [25]



Adaptive Neuro-Fuzzy inference systems (ANFIS)
for MR brain tumor classification developed in this
research and a comprehensive feature set and fuzzy
rules are selected to classify an abnormal image to
the corresponding tumor type
The classification accuracy of ANFIS is
comparatively higher than the fuzzy and neural
classifiers.
The convergence time period of ANFIS is ten times
better than the neural and the fuzzy classifier.
(DWT) principal component analysis (PCA),
(BPANN) k nearest neighbor classifier (KNN) [26]
4.5.2

classify subjects as normal or abnormal MRI human
images,

A classification with a success of 97% by FP-ANN
and 98% has been obtained by k-NN,

This research developed two hybrid techniques,
DWT + PCA + FP-ANN and DWT + PCA +k-NN
to classify the human brain MR images. Without
any tumor detection.
4.5.3 Neural network, adaptive adjustment [27]


The EM is initially used in brain MR image
histogram for estimating the distribution of
parameters for WM, GM and CSF regions.
These estimated parameters work as fitness function
for tissue classification.
5. CONCLUSION
Proper segmentation method is required for accurate diagnosis
of brain tumor detection and segmentation. Currently various
slices of many images provide information for planning and
treatment purpose. In this paper, we reviewed some of recent
research work done on brain tumor detection and
segmentation. By analyzing literature we conclude that the
automation of brain tumor detection and segmentation using
MRI is one of the most active research area but currently there
is no method which is accepted clinically.
6. REFERENCES
[1] Ferlay J, Shin HR, Bray F, Forman D, Mathers C and
Parkin DM, GLOBOCAN 2008 v2.0, Cancer Incidence and
Mortality Worldwide, International Agency for Research on
Cancer, Lyon, France, 2010. http://globocan.iarc.fr, Accessed
on: November 13, 2011.
[2] Louis D.N., Ohgaki H., Wiestler O.D, Cavenee W.K.
(Eds.), WHO Classification of Tumors of the Central Nervous
System, International Agency for Research on Cancer
(IARC), Lyon, France, 2007.
[3] http://www.radiologyassistant.nl/, Accessed on: January,
12, 2012.
[4] A. O Rodriguez, Principles of Magnetic Resonance
Imaging, Revista Mexicana de Fisica, Vol. 50, No. 3, 2004,
pp. 272-286.
[5] Jain R. et al, Machine Vision, McGraw-Hill, Inc. 1995.
[6] Sonka, M., Hlavac, V., and Boyle, R., “Image Processing,
Analysis, and Machine Vision”, Brooks / Cole Publishing
Company, 1998.
[7] Zucker, S. W., “Region growing: Childhood and
adolescence”, Computer Graphics and
Image Processing, 1976, 5, 382-399.
[8] Sezgin M. and Sankur B., Survey over image thresholding
techniques and quantitative performance evaluation, Journal
of Electronic Imaging, Vol. 13, No. 1, Jan. 2004, pp. 146–
165.
4
[9] K. K. Singh, A. Singh, “A Study of Image Segmentation
Algorithms for Different Types of Images”, International
Journal of Computer Science Issues, Vol. 7, Issue 5, 2010.
images." Computer Engineering and Systems (ICCES), 2010
International Conference on.IEEE, 2010.
[10] W. X. Kang, Q. Q. Yang, R. R. Liang, “The Comparative
Research on Image Segmentation Algorithms”, IEEE
Conference on ETCS, pp. 703-707, 2009.
[21] Zhang, Yudong, Zhengchao Dong, Lenan Wu, and
Shuihua Wang. "A hybrid method for MRI brain image
classification." Expert Systems with Applications38, no. 8
(2011): 10049- 10053.
[11] Zhang, Y. J, An Overview of Image and Video
Segmentation in the last 40 years, Proceedings of the
6thInternational Symposium on Signal Processing and Its
Applications, pp. 144-151, 2001.
[22] Dou, W., Ruan, S., Chen, Y., Bloyet, D., & Constans, J.
M. (2007). A framework of fuzzy information fusion for the
segmentation of brain tumor tissues on MR images. Image
and vision Computing, 25(2), 164-171.
[12] J. C. Dunn, A Fuzzy Relative of the ISODATA Process
and Its Use in Detecting Compact Well-Separated Clusters,
Journal of Cybernetics, Vol. 3, No.3, 1973, pp. 32-57.
[23] Sikka, Karan, Nitesh Sinha, Pankaj K. Singh, and Amit
K. Mishra. "A fully automated algorithm under modified
FCM framework for improved brain MR image
segmentation." Magnetic Resonance Imaging 27, no. 7
(2009): 994-1004.
[13] Dehariya, Vinod Kumar, Shailendra Kumar Shrivastava,
and R. C. Jain. "Clustering of Image Data Set Using K-Means
and Fuzzy KMeans Algorithms."Computational Intelligence
and Communication Networks (CICN), 2010 International
Conference on. IEEE, 2010.
[14] M. Arfan Jaffar, Quratulain, and Tae Sun Choi, Tumor
Detection From Enhanced Magnetic Resonance Imaging
Using Fuzzy Curvelet, Microscopy Research and Technique,
Wiley Online Library, Vol. 75, No. 6, April 2012, pp. 499504.
[15] Nandita Pradhan and A.K. Sinha, Fuzzy ANN Based
Detection and Analysis of Pathological and Healthy Tissues in
FLAIR Magnetic Resonance Images of Brain, international
Journal of Information Technology and Knowledge
Management, Vol. 4, No. 2, July- Dec. 2011, pp. 471-476.
[16] Prastawa, Marcel, and Guido Gerig. "Automatic MS
lesion segmentation by outlier detection and information
theoretic region partitioning." Grand Challenge Work.: Mult.
Scler.Lesion Segm. Challenge (2008): 1-8.
[24] Van Lung, H., & Kim, J. M. (2009, August).A
generalized spatial fuzzy c-means algorithm for medical
image segmentation. In Fuzzy Systems, 2009. FUZZ - IEEE
2009. IEEE International Conference on (pp. 409-414). IEEE.
[25] Hemanth, D. Jude, C. K. Vijila, and J. Anitha.
"Application of Neuro-Fuzzy Model for MR Brain Tumor
Image Classification." Int J Biomed Imaging 16 (2009): 95102.
[26] El-Dahshan, El-Sayed Ahmed, Tamer Hosny, and AbdelBadeeh M. Salem. "Hybrid intelligent techniques for MRI
brain images classification." Digital Signal Processing 20.2
(2010): 433-441.
[27] Fu, J. C., Chen, C. C., Chai, J. W., Wong, S. T., & Li, I.
C. (2010). Image segmentation by EM-based adaptive pulse
coupled neural networks in brain magnetic resonance
imaging. Computerized Medical Imaging and Graphics, 34(4),
308-320.
[17] Dubey, R. B., Hanmandlu, M., Gupta, S. K., & Gupta, S.
K. (2009). Semi-automatic segmentation of MRI Brain tumor.
ICGSTGVIP Journal, 9(4), 33-40.
[18] Singh, Laxman, R. B. Dubey, Z. A. Jaffery, and Z.
Zaheeruddin. "Segmentation and Characterization of Brain
Tumor from MR Images". In Advances in Recent
Technologies in Communication and Computing, 2009.
ARTCom'09. International Conference on, pp. 815-819.
IEEE, 2009.
[19] Amruta, A., Abhijeet Gole, and Yogesh Karunakar. "A
systematic algorithm for 3-D reconstruction of MRI based
brain tumors using morphological operators and bicubic
interpolation." Computer Technology and Development
(ICCTD), 2010 2nd International Conference on. IEEE, 2010.
[20] Badran, Ehab F., Esraa Galal Mahmoud, and Nadder
Hamdy. "An algorithm for detecting brain tumors in MRI
5
Download