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

Volume: 4 Issue: 3 367 – 372

_______________________________________________________________________________________________

Enhancement Techniques and Methods for Brain MRI Imaging

Sukhjeet Kaur
Research Scholar
Punjabi University
Patiala, India
Sukhygill35@gmail.com

Abstract— In this paper, it is planned to review and compare the different methods of enhancing a DICOM of brain MRIused in preprocessing
and segmentation techniques. Image segmentation is the process of partitioning an image into multiple segments, so as to change the
representation of an image into something that is more meaningful and easier to analyze. Several general-purpose algorithms and techniques
have been developed for image segmentation. This paper describes the different segmentation techniques used in the field of ultrasound, MR
image and SAR Image Processing. In preprocessing and enhancement stage is used to eliminate the noise and high frequency components from
DICOM image. In this paper, various Preprocessing and Enhancement Technique, Segmentation Algorithm and their compared.

Keywords—Enhancement, Pre-processing, Medical-Imaging, Image Segmentation, Center Weighted Median (CWM) Filter.

___________________________________________*****______________________________________________

I. INTRODUCTION and analysis, aiming to improve the appearance of image in

The image processing techniques, image enhancement terms of human brightness perception. Contrast enhancement
techniques and de- noising techniques are widely used in the is among them and is often part of image processing systems
field of medical imaging, especially Magnetic Resonance in the pre-processing and/or post-processing stage.
Imaging (MRI). The test that makes use of magnetic field and
radio pulses to capture images of body especially that of the II. IMAGE ACQUISITION
brain is called Magnetic Resonance Imaging. The MRI is To access the real medical images for carrying our research is
superior to the X- ray and Computed Tomography or CT scan, a very difficult because of privacy issues and heavy technical
as it can provide information of body tissues that cannot be difficult. This idea is automatic brain tumor detection methods
provided by the others. Basically, the MRI technique of through MRI brain Images. A sample of 80 T1 weighted
imaging is employed in medicinal backgrounds to generate images is used for enhancement purpose.T1- weighted images
high quality images of the human body parts. The MRI works visible water darker and the fat brighter. All MRI images were
on the principle of Nuclear Magnetic Resonance (NMR), acquired on a 0.5T open prevent MRI system [2]. Most
which is a spectroscopic technique used by thescientists to Medical Imaging Studies and detection are directed using
acquire microscopic information about the molecules. The MRI, Positron Emission Tomography (PET) and Computed
MRI is now advanced and uses the volume imaging technique tomography (CT) Scan.
and Magnetic Resonance Imaging is young and is growing in
the field of science. The most suitable technique for the
problemsof soft tissues differentiates MRI from other imaging
variants in the medicinal field. An x- ray is ideal for hard
tissue like bones while a MRI is ideal for soft tissue like
human brain. The MRI is used for a number of purposes like
diagnosing braintumors, multiple sclerosis, and spinal
infections; to visualize torn ligaments, shoulder injuries,
tumors in bones, herniated discs and strokes at initial stages.
The MRI images clearly show the tissues that are damaged or
diseased and these references. In order to view the MRI Figure 1: Sample of Brain MRI
images clearly, sometimes a contrast material is also used and
it helps to identify flow of blood, tumors and inflamed areas in III. PREPROCESSING TECHNIQUES
the tissues. To give proper diagnosis and good results, doctors A. Removal of film artifacts
are provided with the different results of enhanced images. The MRI brain image consists of film artifacts or label on the
Enhancement is a fundamental task in digital image processing MRI include patient name, age and marks. Modify
367
IJFRCSCE | March 2018, Available @ http://www.ijfrcsce.org
_______________________________________________________________________________________
International Journal on Future Revolution in Computer Science & Communication Engineering ISSN: 2454-4248
Volume: 4 Issue: 3 367 – 372
_______________________________________________________________________________________________
trackingalgorithm used to eliminate film artifact. This IV. ENHANCEMENT
algorithm analysis the brain tumor of image the first row and Image enhancement techniques used to develop the visual
first column is started, the intensity value of the pixels are appearance of images from Magnetic Resonance Image (MRI)
analyzed and found the threshold value of the film and the enhancing brain volumes were aligned linear and this
image is contrast. The enhancement operate are eliminate of
B. Removal of Skull using Modified Tracking Algorithm film artifacts and labels, filtering of images. There are more
To eliminate unwanted portion of MRI by using modified number of enhancement techniques such as Median filter,
Tracking Algorithm. The MRI brain image is left, right and Gabor Filter, Gaussian Filter, low pass filter and Prewitt edge-
top skull portions. To obtain finally eliminating the film finding filter.
artifacts and labels is taken. Brain imaging is started from first
row, first column of left the given matrix and select the peak V. SEGMENTATION
threshold value from matrix of left side and considers the 200 Segmentation is the process of partitioning a digital image into
flag value. segments i.e. into small number of pixels to simplify
therepresentation of the image. There are several segmentation
C. Need for pre-processing methods. Thresholding technique in which input gray scale
The pre-processing is required for MRI images as: image is converted into binary image based on some threshold
value. Thersholding is of two types a) Global b) Local In
∑ Marks or labels present (Film artifacts) can prevent in the watershed transformations pixels of an image are grouped on
post-processing of these images. the basis of their intensities. Morphological operations are
∑ Images are to be made more suitable for advanced non-linear operations which are related to the shape and
processing in CAD systems. morphology of features in animage. K- means clustering is an
∑ Enhance the quality of image. algorithm which is used togroup pixels based on attributes into
∑ Remove the noise of image. k number of groupswhere k is a positive integer. Fuzzy c-
means is a several authors suggested various algorithms for
Preprocessing is used to displaying the digital images segmentation wayof processing the data by giving the partial
orhighlighting edges of image. The Preprocessing membership value to each pixel in the image. Genetic
Techniquesinclude the Content Based model, Fiber tracking algorithm is basedon heuristic method. It works in five stages
Method, Wavelets& Wavelet Packets, and Fourier i.e.
transformtechnique.
1. Initialization of population
TABLE I : AN OVERVIEW OF PREPROCESSING 2. Evaluation of fitness function
TECHNIQUES 3. Selection
METHODS REMARKS 4. Crossover
5. Mutation and termination.
Data is large
Neural Networks, Genetic
volume the processed
Programming[62] Inneutrosophic theory, every event has not only a
successfully.
certaindegree of the truth, as well as a falsity degree and an
Head Model, Finite
It is used to analyze the indeterminacy degree that must be considered
DifferenceTime-DomaiN
more type of tissue. autonomouslyfrom one another. Region growing technique is
(FDTD)[59]
pixel basedimage segmentation. In normal region growing
To reduce the artifacts
PCA (Principal Component method,only the intensity constrain is taken into account.
present in the PET data
Analysis)[60] Here, firsta threshold value is set and a seed point is found out.
set.
Theneighboring pixels, whose intensity difference between
Pixel It removes the noise and it
theseed point and corresponding neighbor pixel is below
Histograms,Morphological can improve the integrity
Process[63] performance. theparticular threshold value, are grown to the region. [38, 39]

The satisfied of MR-DT1

Fiber tracking Method, A. Literature Survey
datasets.
Runge-Kutta method[61] G. Evelin et al, 2013; presented threshold based
It is use to image of imagesegmentation. The outcome shows the proper detection
Geometric prior, Bimodal[58] ofregion of interest.
registered.
It represents the idealized
Boundary Model ,Nonlinear
MR intensity profile
matching scheme[64]
accurately.
368
IJFRCSCE | March 2018, Available @ http://www.ijfrcsce.org
_______________________________________________________________________________________
International Journal on Future Revolution in Computer Science & Communication Engineering ISSN: 2454-4248
Volume: 4 Issue: 3 367 – 372
_______________________________________________________________________________________________
Jin Liu et al, 2014; presented a review paper on Yongyue Zhang et al, 2001; proposed a popularstochastic
differentsegmentation methods. The results of this paper show model provides the better framework for manycomplex
the comparison of different segmentation techniques. problems in image segmentation is MarkovRandom Field
(MRF) model [4]. MRF model and itsvariants have been
Nikhil R. Pal et al, 1993; presented image successfully used for brain MR imagesegmentation.
segmentationchallenging problems that has received an
enormous amount of attention by many researchers. S. Ruan et al. proposed a fuzzy Markova method for brain
Authorspresented various techniques used in medical tissue segmentation from magnetic resonance images that
imagesegmentation and analysis. The segmentation problem calculates fuzzy membership in each pixel to indicate the
canbe categorized as supervised and unsupervised problem. partial volume degree, which is statistically modeled.
For appropriate analysis, different image models have
beenproposed for taking care of spatial intrinsic
characteristics.
TABLE II: A BRIEF SURVEY OF SEGEMENTATION METHODS

Year of
Author Name Method Used Remarks
Publication
Linde Buzo-Gray algorithm (LBG) used for
2009 H.B et al LindeBuzo-Gray algorithm method[40]
segmentation of MRI images.
AC entirely on the MRI data obtained and used
2010 A. Gregory et al Attenuation- Correction Method[37] for neurologic performed with the MR-PET
human brain scanner prototype.
Unsupervised MR image Segmentation Fuzzy C-mean clustering algorithm for
2010 T.Logeswari et al
method[46] Segmentation is presented.
To achieve lowest value of weight vector, a
T.Logeswari et al
Hierarchical-Self Organizing-Map highest value of tumor pixels, computation speed.
2010
(HSOM)[38] It’s used to classify the brain tumor of MRI
image.
To PET attenuation correction (AC), used two
Matthias et al Atlas registration and pattern algorithms for whole-body MRI-based AC
2011
recognition[41] (MRAC) based on atlas registration and pattern
recognition (AT&PR)
To assess brain gray (GM) and white matter
Yu Zhang et al
Alzheimer's-disease (AD) and front (WM) abnormalities jointly between the
2011
temporal Dementia (FTD)[42] disease to elucidate differences in abnormal MRI
image.
Yannick et al Ultra short-Echo- Time/ Dixon MRI To separate cortical bone and air, the Dixon
2012
sequence[44] technique for soft and adipose tissues are used.

Matthew et al
To develop and validate MRI methods for
2012 Momestic pig method[41]
estimating brain volume of image.

It using 3D statistical features extracted from the

SOM-FCM-Based Method and 3D
2013 Andres et al brain tumor image are proposed effective and
Statistical.[24]
efficient of segment of image
It used to efficiently pixels of image as belonging
2013 Francisco et al Gaussian Mixture Model[45] to either the low-intensity or high-intensity
background of image.

369
IJFRCSCE | March 2018, Available @ http://www.ijfrcsce.org
_______________________________________________________________________________________
International Journal on Future Revolution in Computer Science & Communication Engineering ISSN: 2454-4248
Volume: 4 Issue: 3 367 – 372
_______________________________________________________________________________________________
Zhang et al. proposed Hidden Markov Random Field the construct the initial membership matrix to clusters into
(HMRF) model to achieve brain MR image segmentation in improve the strength of the clusters [20].
unsupervised framework [6]. The segmentation obtained
choice of initial model parameters. As Expectation-
VI. CENTER WEIGHTED MEDIAN FILTER
Maximization algorithm yield solutions at the cost of high
Median filter is used to eliminate the noise and high
computational burden, in order to overcome this Marroquin
frequency components from MRI image. and reduce’ salt
et al. have proposed a new class of probabilistic model,
and pepper’ noise and without disturbing the edges. A
called
median is calculated by all pixel values sorted by their size,
Hidden Markov Measure Field model that solved the
then selecting the new value as median value for the pixel.
complex segmentation problem by minimization of
The amount of 3*3 window pixels should be used to
differentiable energy function.
calculate the median. A weighted median filter used for
removing noise from MRI brain images with contrast. It has
Wellsetal et al. proposed an adaptive brain MR image
a great potential for being used in rank order filtering and
segmentation scheme in EM framework [19,20].They have
image processing. The Center Weighted Median (CWM)
also taken spatial intensity in-homogeneity into account and
filter, is a weighted median filter .it uses to weight of value
Have estimated the bias field. Recently Hung et al. proposed
of each window. This filter can preserve image details while
an automatic segmentation method based on a decision tree
suppressing additive white and/or impulsive-type noise. The
to different classes the brain tissues in magnetic resonance
statistical properties of the CWM filter are analyzed.
(MR) images.
VII. CONCLUSIONS
Martin et al. Described away to segment the medical
In this review paper, various method of Image Segmentation
images using an appropriately defined fuzzy clustering
through MRI are studied and compared for preprocessing
based on a fuzzy relation. The considered relation is defined
and segmentation techniques. It is used to give large
in terms of Euclidian distance.
information about DICOM image segmentation and
detection and milestones for analyze from different MRI
Panas et al. proposed the Adaptive Fuzzy
medical image processing. In this paper, various steps in
Clustering/Segmentation (AFCS).In AFCS, then on-
Preprocessing and Enhancement Technique, Segmentation
stationary nature of the taken the image in account using
Algorithm and their performance have been studied and
modifying the prototype vectors as function of sample
compared.
location in the image. A multimodal is utilized for varies of
estimating the spatially prototype vectors for different
REFERENCES
window sizes. The results provide segmentation having
[1] G. Evelin Sujji, Y.V.S. Lakshmi, G. Wiselin Jiji (2013),
lower entropy. Mohamed et al. described the application of MRI Brain Image Segmentation based on Thresholding,
fuzzy settheory in medical imaging International Journal of Advanced Computer Research
Guanetal et al. have proposed an automatic hotspot (Volume-3 Number-1 Issue-8)
detection and segmentation of whole body PET images [2] Jin Liu, Min Li, Jianxin Wang_, Fangxiang Wu, Tianming
usingthreshold and the Hidden Markov Model(HMM) .They Liu, and Yi Pan (2014), A Survey of MRI-Based Brain
compare the fixed PET pixel data threshold and the Tumor Segmentation Methods, Tsinghua Science and
fixedstandard up take values(SUV) threshold for segmenting Technology, pp578-595 Volume 19, Number 6
[3] Nikhil R. Pal and Sankar K. Pal, “A review on image
hotspots.
segmentation tech- niques,”Pattern Recognition, vol. 26,
no. 9, pp. 1277-1294, Sept.1993
Nanda et al. proposed a Tabu search based [4] Young yue Zhang, Michael Brady and Stephen Smith,
unsupervisedscheme using HMRF-EM framework which “Segmentation of Brain MR Images through a Hidden
could segmentthe images properly taking arbitrary initial Markov random Field Model and the Expectation-
parameter.To obtain cluster is proposed of fully automatic Maximization Algorithm,” IEEE Trans. Medical Imaging,
technique.A modified fuzzy c-means classification vol.20, no.1, pp. 45-57.
algorithm is by to provide a fuzzy partition. The method is [5] J. L. Marroquin, E. A. Santana and S. Botello, “Hidden
inspired byMarkov random field (MRF) and to be less Markov Measure Field Models for Image segmentation,”
IEEE Trans. Pattern Analysis and Machine Intelligence,
sensitive to noise found as it filters the image while
vol.25, no.11, 1380-1387, 2003.
clusterin Karnan et al. presented a new method called
[6] J. Zhang, J.W. Modestino and D.A. Langan, “Maximum
Fuzzy Membership C-Means (FMCM) for segmentation Likelihood parameter estimation for unsupervised
ofMagnetic Resonance Images (MRI). This work develops stochastic model-based image segmentation,” IEEE Trans.
Image Processing, vol.3, no. 4, pp.404-420,1994.
370
IJFRCSCE | March 2018, Available @ http://www.ijfrcsce.org
_______________________________________________________________________________________
International Journal on Future Revolution in Computer Science & Communication Engineering ISSN: 2454-4248
Volume: 4 Issue: 3 367 – 372
_______________________________________________________________________________________________
[7] P. K. Nanda, “MRF model learning and application to Segmentation”. Signal Acquisition and Processing, 2010.
image restoration and segmentation,” Ph.D Dissertation, ICSAP '10. International Conference.
IIT Bombay, 1995. [23] H.B. Kekre, saylee M. Gharge, “Image segmentation of
[8] W. M. Wells, W. E. L. Grimson, R. Kikinis, F.A. Jolesz, MRI images using vector quantization
“Adaptive Segmentation of MRI Data,” IEEE Trans. techniques”.Springer link.
Medical Imaging, vol. 15,pp. 429-442, Aug. 1996. [24] Yannick Berker, Jochen Franke, André Salomon, “MRI-
[9] K. Held, E. R. Kops, B. J. Krause, W. M. Wells and R. Based Attenuation Correction for Hybrid PET/MRI
Kikinis, H.W. M. Gartner, “Markov Random Field Systems. A 4- ClassTissue Segmentation Technique Using
Segmentation of Brain MRI Images,” IEEE Trans. Medical a Combined Ultra short-Echo-Time/Dixon MRI
Imaging, vol. 16, pp. 878-886, Dec.1997. Sequence”. 2012 vol. 53 no. 5 796-804.
[10] WenHung Chaoa, You-Yin Chena, Sheng-Huang Linc, [25] Ajala Funmilola, Oke O.A, Alade O.M, Adewusi E.A
Yen Yu I. Shihd, Siny Tsange , “Automatic segmentation “Fuzzy kc- means Clustering Algorithm for Medical Image
of magnetic resonance images using a decision tree with Segmentation”.ISSN Paper 2224-5782 ISSN Online 2225-
spatial information,”Computerized Medical Imaging and 0506.
Graphics vol.33, pp. 111-121, 2009. [26] Francisco Gimenez, Adrian P. Kells, “Automated
[11] Thakur and R.S. Anand, “Image Quality Based Segmentation Tool for Brain Infusions”, Featured in PLOS
Comparative Evaluation of Wavelet Filters in Ultra sound Collections, 2013.
Speckle Reduction,” Digital Signal Processing, Vol.15, [27] Paul Schmidt, Dorothea Buck, Susanne Bührlen, Mark
No.5, pp 455-465, 2005. Mühlau, “Application to Segmentation and
[12] H. Guan, T. Kubota, X. Huang, X. S. Zhou, and M.Turk. StatisticalAnalysis of T2-
“A1utomatic hot spot detection and segmentation in whole Hypointensitie”.Jul17,2013DOI:10.1371/journal.pone.0
body fidget images,” Proceedings of IEEE International 068196 Featured in PLOS Collections.
Conference on Image Processing (ICIP), 2006. [28] Moritz F Kircher, Eric C Holland, “A brain tumor
[13] M.Y. Siyal and Lin Yu, “An intelligent modified fuzzy c- molecular imaging strategy using a new triple modality
means based algorithm for bias estimation and MRI-photo acoustic Raman nano particle”. Nature
segmentation of brain MRI,” Pattern Recognition letters, Medicine 18, 829–834 (2012) doi: 10.1038/nm.
vol.26, pp. 2052-2062, 2005. [29] Ying Wang, Zhi Xian Lin, Mao Qing Li, “Automatic MRI
[14] Ping Wang, Hong Lei Wang, “A modified FCM algorithm Brain Tumor Segmentation System Based onLocalizing
for MRI brain image segmentation,” IEEE Trans. on Future Active ContourModels”.
Bio-Medical Information Engineering, pp.26-29, 10.4028/www.scientific.net/AMR.219-220.1342.
Dec.2008.. [30] E. Ben George, M. Karnan, “MR Brain Image
[15] Aboul Ella Hassanien, “Fuzzy rough sets hybrid scheme Segmentation using Bacteria Foraging Optimization
for breast cancer detection,” Image and Vision Computing, Algorithm. International Journal of Engineering and
vol.25, pp. 172- 183, 2007. Technology (IJET)”
[16] Tabakov Martin, “A fuzzy clustering technique for medical [31] T. Logeswari, Karnan M. “An Enhanced Implementation
image segmentation”, Proceedings of internal symposium of Brain Tumor Detection Using Segmentation Based on
on evolving fuzzy systems, pp. 118-122, sept.2006. Soft Computing”. Signal Acquisition and Processing, 2010.
[17] K. Jiayin, Min Lequan, “Novel modified fuzzy c-means ICSAP '10. International Conference.
algorithm with applications,” Digital signal processing, vol. [32] Karnan M, Gopal N.N, “Hybrid Markov Random Field
19, no.2, pp. 309- 319, Mar, 2009. with Parallel Ant Colony Optimization and Fuzzy C Means
[18] Tolias Yannis and M. Panes Stavros M, “Image for MRI Brain Image segmentation”. Computational
segmentation by a fuzzy clustering algorithm using Intelligence and Computing Research (ICCIC), 2010 IEEE
adaptive spatially constrained functions,” IEEE International Conference.
Transactions on systems, Man, and Cybernetics vol.28, [33] T.Logeswari and M.Karnan, “An Improved
issue.3, pp. 359-369, May, 1998. Implementation of Brain Tumor Detection Using
[19] N.A, Mohamed, M.N. Ahmed, “Modified fuzzy c-means in Segmentation Based on Hierarchical Self Organizing
medical image segmentation,” Proceedings of IEEE Map”. International Journal of Computer Theory and
international conferenceon Acoustics, speech, and signal Engineering, Vol. 2, No. 4, August, 2010 .1793-8201.
processing, vol. 6, pp. 3429-3432, 1999. [34] Thomas M Hsieh,Yi-Min Liu, “Automatic segmentation of
[20] Kannan S.R., “A new segmentation system for MR images meningioma from non-contrasted brain MRI integrating
based on fuzzy techniques,” Applied soft computing, vol. fuzzy clustering and region growing”. BMC Medical
8, issue. 4, pp. 1599-1606, Sept. 2008. Informatics and Decision Making 2011,11:54.
[21] S. Ruan, B. Moretti, J. Fadili and D. Bloyet, “Fuzzy [35] Xiaofeng Yang, Baowei Fei, “Multi scale segmentation of
Markovian segmentation in application of magnetic the skull in MR images for MRI-based attenuation
resonance images,” Computer Vision and Image correction of combined MR/PET”. Jamia.bmj.com J Am
Understanding. vol. 85, pp. 54-69, 2002. Med Inform Assocdoi: 10.1136/amiajnl-2012-001544.
[22] T. Logeswari, M. Karnan, “Hybrid Self Organizing Map [36] Miguel Angel, “Structural MRI in Fronto temporal
for Improved Implementation of Brain MRI Dementia:Comparisons between Hippo campal Volumetry,
Tensor Based Morphometry and Voxel Based
371
IJFRCSCE | March 2018, Available @ http://www.ijfrcsce.org
_______________________________________________________________________________________
International Journal on Future Revolution in Computer Science & Communication Engineering ISSN: 2454-4248
Volume: 4 Issue: 3 367 – 372
_______________________________________________________________________________________________
Morphometry”. Dec 20, 2012 DOI:
10.1371/journal.pone.0052531.
[37] M. Carmen, Randolph J. Nudo, PhD. “Primary Motor
Cortex in Stroke A Functional MRI-Guided Proton MR
SpectroscopicStudy”.stroke.ahajournals.org.February17,
2011,doi:10.1161/STRO KEAHA.110.601047.
[38] C. Engineering, “Enhancement Techniques and Methods
for MRI- A Review,” vol. 5, no. 1, pp. 397–403, 2014.
[39] G. Kaur and J. Rani, “MRI Brain Tumor Segmentation
Methods- A Review,” vol. 6, no. 3, pp. 760–764, 2016.
[40] H.B. kekre, Sayleem Gharge, “Detection and Demarcation
of Tumor using Vector Quantization in MRI images”. 23
Jan 2010 IJEST Volume 1 Issue 2 2009 59-66.
[41] Matthias Hofmann, Frederic Mantlik, “MRI-Based
Attenuation Correction for Whole-Body PET/MRI:
Quantitative Evaluation of Segmentation and Atlas-Based
Methods”, 2011 vol. 52 no. 9 1392- 1399.
[42] Andrew J. Patterson, Victoria E. Young, Jonathan H.
Gillard, “An Objective Method to Optimize the MR
Sequence Set for Plaque Classification in Carotid Vessel
Wall Images Using Automated Image Segmentation”. Oct
23, 2013 DOI: 10.1371/journal.pone.0078492.
[43] T. Logeswari, M. Karnan, “Hybrid Self Organizing Map
for Improved Implementation of Brain MRI
Segmentation”. Signal Acquisition and Processing, 2010.
ICSAP '10. International Conference.
[44] Yannick Berker, Jochen Franke, André Salomon, “MRI-
Based Attenuation Correction for Hybrid PET/MRI
Systems. A 4- ClassTissue Segmentation Technique Using
a Combined Ultra short-Echo-Time/Dixon MRI
Sequence”. 2012 vol. 53 no. 5 796- 804.
[45] Francisco Gimenez, Adrian P. Kells, “Automated
Segmentation Tool for Brain Infusions”, Featured in PLOS
Collections, 2013.

372
IJFRCSCE | March 2018, Available @ http://www.ijfrcsce.org
_______________________________________________________________________________________