Brain Tumor Detection Through MR Images: A Review of Literature
Brain Tumor Detection Through MR Images: A Review of Literature
Brain Tumor Detection Through MR Images: A Review of Literature
e-ISSN: 2278-0661,p-ISSN: 2278-8727, Volume 17, Issue 5, Ver. II (Sep. – Oct. 2015), PP 07-18
www.iosrjournals.org
Abstract: A brain tumor is an abnormal growth of tissue in the brain or central spine that can disrupt proper
brain function and creates an increasing pressure in the brain. This paper is intended to present a
comprehensive review of the methods of brain tumor detection through Magnetic Resonance Imaging (MRI)
technique used in different stages of Computer Aided Detection System (CAD). It also provides a brief
background on brain tumor in general and non-invasive imaging of brain tumor in order to give a
comprehensive insight into the field. Lastly, the paper concludes with a concise discussion and provides a
direction toward the upcoming trend of more advanced research studies on brain image segmentation and
tumor detection.
Keywords: Brain Tumor, Central Nervous System (CNS), Cerebrospinal Fluid (CSF), Magnetic Resonance
Imaging (MRI), Segmentation.
I. Introduction
A brain tumor is a collection (or mass) of abnormal cells in the brain. A tumor may lead to cancer,
which is a major leading cause of death and responsible for around 13% of all deaths world-wide. Cancer
incidence rate is growing at an alarming rate in the world. So detection of the tumor is very important in earlier
stages. Great knowledge and experience on radiology are required for accurate tumor detection in medical
imaging. MRI is the most flexible of our diagnostic imaging modalities, possessing the ability to characterize a
wide range of parameters in the living subject and provide exquisite spatial resolution. Brain tumor
identification form magnetic resonance imaging (MRI) consists of several stages. Segmentation is known to be
an essential but difficult step in medical imaging classification and analysis. Hence, it is highly necessary that
segmentation of the MRI images must be done accurately before asking the computer to do the exact diagnosis.
This review presents an overview of magnetic resonance imaging (MRI)-based medical image analysis for brain
tumor studies.
1.1 Brain
Together, the brain and spinal cord (the central nervous system (CNS)) control the physiological and
psychological functions of our body. Generally our brain includes three major parts:
1. Cerebrum. It controls thinking, learning, problem solving, emotions, speech, reading, writing, and voluntary
movement.
2. Cerebellum. It controls movement, balance, and posture.
3. Brain stem. It connects the brain to the spinal cord, and controls vital functions in human body such as
motor, sensory pathways, cardiac, repository and reflexes [1].
The brain is composed of two tissue types, namely gray matter (GM) and white matter (WM). Gray
matter is made of neuronal and glial cells, also known as neuroglia or glia that controls brain activity and the
basal nuclei which are the gray matter nuclei located deep within the white matter. The basal nuclei include:
caudate nucleus, putamen, pallidum and claustrum. White matter fibers consist of many elinated axons which
connect the cerebral cortex with other brain regions. The left and the right hemispheres of the brain are
connected by corpus callosum which is a thick band of white matter fibers. Both, cerebellum and cerebrum
have a very thin outer cortex of gray matter, internal white matter and small but deeply situated masses of the
gray matter. The spinal cord is located toward the bottom of the brain. It has three structures: the midbrain, pons
and medulla oblongata [2].
The brain also contains cerebrospinal fluid (CSF) which consists of glucose, salts, enzymes, and white
blood cells. This fluid circulates through channels (ventricles) around the brain and the spinal cord to protect
them from injury. There is also another tissue called meninges which are the membrane covering the brain and
spinal cord [2].
Fig. 1 [3] shows the anatomy of the brain. It is composed of the cerebrum and the brain stem. The
cerebrum occupies the
Largest part of the brain. It is connected with the conscious thoughts, movement and sensations. It
further consists of two halves, the right and the left hemispheres. Each controls the opposite side of the body.
Moreover, each hemisphere is divided into four lobes: the frontal, temporal, parietal and occipital lobes as
shown in Fig. 2 [3].
from different cell types, and may have different treatment options. Scientists have classified brain tumor
according to:
1. The type and grade (how aggressive it is),
2. Whether it is a primary or a secondary tumor,
3. If it is cancerous (malignant) or not (benign), and
4. Where in the brain the tumor is located [5].
The least aggressive type of brain tumor is often called a benign brain tumor. They originate from cells
within or surrounding the brain, do not contain cancer cells, grow slowly, and typically have clear borders that
do not spread into other tissue. They may become quite large before causing any symptoms. If these tumors can
be removed entirely, they tend not to return. Still, they can cause significant neurological symptoms depending
on their size, and location near other structures in the brain. Some benign tumors can progress to become
malignant.
Malignant brain tumors contain cancer cells and often do not have clear borders. They are considered
to be life-threatening because they grow rapidly and invade surrounding brain tissue. Although malignant brain
tumors very rarely spread to other areas of the body, they can spread throughout the brain or to the spine. These
tumors can be treated with surgery, chemotherapy and radiation, but they may recur after treatment.
Whether cancerous or benign, tumors that start in cells of the brain are called primary brain tumors. Primary
brain tumors may spread to other parts of the brain or to the spine, but rarely to other organs. Metastatic or
secondary brain tumors begin in another part of the body and then spread to the brain. These tumors are more
common than primary brain tumors and are named by the location in which they begin. They are treated based
on where they originate, such as the lung, breast, colon or skin. Each of these tumors has unique clinical,
radiographic and biological characteristics [4].
The pituitary gland is located at the base of the brain and it produces hormones that control other
glands in the body; specifically the thyroid, adrenal glands, ovaries and testes, glands for milk production in
pregnant women, and the kidneys. Tumors in or around the pituitary gland can lead to problems with how these
glands function. Also, patients may have vision problems. Pituitary tumors are frequently benign, and surgical
removal is often the cure. Some are treated with medication to shrink or stop the tumor from growing [1].
CNS (Central Nervous System) Lymphoma is a malignant primary brain tumor that originates from the
lymphocytes found in the brain, spinal cord, or eyes. It typically remains confined to the CNS. Treatment
commonly includes chemotherapy and/or radiation [1].
An axial MRI looks at the brain from below in a series of images starting at the chin and moving to the
top of the head. A sagittal MRI looks at the brain from the side in a series of images starting at one ear and
moving to the other. A coronal MRI looks at the brain from behind in a series of images starting at the back of
the head and moving to the face as shown in Fig. 4 [10].
a) b) c)
Fig. 4 Brain MR Images from a) Axial Plane, b) Sagittal Plane, and c) Coronal Plane.
The signal processing has three different images that can be achieved from the same body: T1-
weighted, T2- weighted and PD-weighted (proton density) as shown in Fig. 5 [11].
a) b) c)
Fig. 5 Brain MR Images from Axial Plane a) T1-w, b) T2-w, and c) PD-w image.
The signal intensity on the MR image is determined by four basic parameters: 1) Proton density, 2) T1
relaxation time, 3) T2 relaxation time, and 4) Flow. Proton density is the concentration of protons in the tissue in
the form of water and macromolecules (proteins, fat, etc). The T1 and T2 relaxation times define the way that
the protons revert back to their resting states after the initial RF pulse. The most common effect of flow is loss
of signal from rapidly flowing arterial blood [12].
The contrast on the MR image can be manipulated by changing the pulse sequence parameters. A pulse
sequence sets the specific number, strength, and timing of the RF and gradient pulses. The two most important
parameters are the repetition time (TR) and the echo time (TE). The TR is the time between consecutive 90
degree RF pulse. The TE is the time between the initial 90 degree RF pulse and the echo. The most common
pulse sequences are the T1-weighted and T2-weighted spin-echo sequences.
The T1-weighted sequence uses a short TR and short TE (TR < 1000msec, TE < 30msec). The T2-
weighted sequence uses a long TR and long TE (TR > 2000msec, TE > 80msec). The T2-weighted sequence is
usually employed as a dual echo sequence. The first or shorter echo (TE < 30msec) is proton density (PD)
DOI: 10.9790/0661-17520718 www.iosrjournals.org 11 | Page
Brain Tumor Detection through MR Images: A Review of Literature
weighted or a mixture of T1 and T2. This image is very helpful for evaluating periventricular pathology, such as
multiple sclerosis, because the hyperintense plaques are contrasted against the lower signal CSF. More recently,
the FLAIR (Fluid Attenuated Inversion Recovery) sequence has replaced the PD image. FLAIR images are T2-
weighted with the CSF signal suppressed [12].
When reviewing an MR image, the easiest way to determine which pulse sequence was used, or the
"weighting" of the image, is to look at the cerebrospinal fluid (CSF). If the CSF is bright (high signal), then it
must be a T2-weighted imaged. If the CSF is dark, it is a T1-weighted image. Thereafter, look at the signal
intensity of the brain structures. On MR images of the brain, the primary determinants of signal intensity and
contrast are the T1 and T2 relaxation times. The contrast is distinctly different on T1 and T2-weighted images.
Also, brain pathology has some common signal characteristics [12].
Using segmentation in medical images is a very important task for detecting the abnormalities, study
and tracking progress of diseases and surgery planning. Segmentation must not allow regions of the image to
overlap.
Objective of this review section is to present literature survey of image segmentation methods. The
main goal is to highlight advantages and limitations of these methods. Key image processing techniques for
brain MRI image segmentation are classified as thresholding, region-growing, clustering, edge detection, atlas-
based, other methods etc. All these techniques are explained further in the following sections and many notable
methods have been reviewed.
illumination is uneven, the global thresholding is likely to fail. In local thresholding, multiple thresholds are
used to compensate the uneven illumination [20].
Threshold-based: Global Simple and computationally fast. Limited applicability to enhancing tumor
and Local Thresholding areas [35].
Region-based: Region- Simple and capable of correctly segmenting regions that have Partial volume effect [37], [38]. Noise or
growing similar properties and generating connected region [36]. variation of intensity may result in holes or
over-segmentation.
Edge Detection Based Focused on detecting contour. Fail when the image is blurry or too
Method In vision based analysis, edge is considered as a very good complex to identify a given border.
descriptor of contrast [26]. Inability to produce a reasonable solution in
cluttered background.
Watershed Segments multiple regions at the same time, produces a Over-segmentation [40].
complete contour of the images and avoids the need for any
kind of contour joining [39].
Pixel-based: Fuzzy C Unsupervised, always converges the boundaries of tumor. Long computational time, sensitive to noise
Means [41].
Artificial Neural Ability to model non-trivial distributions and non-linear Gathering training samples is not straight-
Networks dependencies [42]. forward and learning phase is slow [43].
Markov Random Fields Are able to represent complex dependencies among data Difficulty when selecting the parameters
instances [44]. that control the strength of spatial
interactions, & usually require algorithms
that are computationally intensive [45].
Model-based: Capable of accommodating to the variability of biological The model may converge to wrong
Parametric Deformable structures over time and across different individuals [46]. boundaries in case of inhomogeneities [47].
Models
Level Sets Approach Topological changes are naturally possible [48]. Computationally expensive [49].
Atlas- Guided Approach Labels are transferred as well as the segmentation. They also Developing the atlas itself is difficult.
provide a standard system for studying morphological
properties, and the data from such study can be used to generate
morphological statistics [34].
M. C. de Andrade, 2004, introduced [51] an interactive algorithm for image smoothing and
segmentation. This method combines some known image smoothing and segmentation methods of mathematical
morphology and PDE-based level set frames. The segmentation was a region growing and automatically detect
all image minima using a property inherited from the watershed transformation (NHW). A merging mechanism
was used to change the image topology which reduces over-segmentation and the need of preprocessing.
Accurate and fast segmentation results were achieved for gray and color images in any number of dimensions
using this method.
Cigdem Demir, et al, 2005, presented [52] a graph-based representation (a.k.a., cellgraph) of
histopathological images for automated cancer diagnosis by probabilistically assigning a link between a pair of
cells (or cell clusters). First, the work defined a set of global metrics on a cell-graph to capture tissue level
information coded into the histopathological images. Second, the results were obtained on the photomicrographs
of 646 archival brain biopsy samples of 60 different patients by comparing the cell-graph approach against cell-
distribution and textural approaches for tissue level diagnosis of brain cancer called malignant glioma. This
method measured the strength of the cell-graph representation which showed 99 percent accuracy for healthy
tissues with lower cellular density level, and at least 92 percent accuracy for benign tissues in the diagnosis of
cancer.
Carles Arus, et al, 2006, introduced [53] HealthAgents, an EC-funded research project to improve the
classification of brain tumors through combination of vivo MRS with in vitro MAS and gene expression.
HealthAgents solved the problem of collection and management of highly complex data by building multi-agent
DOI: 10.9790/0661-17520718 www.iosrjournals.org 15 | Page
Brain Tumor Detection through MR Images: A Review of Literature
decision support over a distributed network of local databases or data marts. They introduced a unique
technology to develop clinical tools for the diagnosis, management and understanding of brain tumors.
G. Farias, et al, 2008, proposed [54] a synergy of signal processing techniques and intelligent strategies was
applied in order to identify different types of human brain tumors, so that it help to confirm the histological
diagnosis. The wavelet-SVM (support vector machine) classifier merged wavelet transform and SVM to reduce
the size of the biomedical spectra and to extract the main features, with SVM to classify them. It reduces the
classification time and improve the results specially taking into account that medical knowledge was not
considered.
Rajeev Ratan, Sanjay Sharma, S. K. Sharma, 2009, have developed [55] a brain tumor segmentation
method on 2D MRI Data which automatically identifies tumor tissue. The watershed segmentation method did
not require any initialization inside the tumor and the visualization and quantitative evaluations of the
segmentation results demonstrate the effectiveness of this approach. This method performance is better for the
cases where the intensity level difference between the tumor and non tumor regions is higher.
Sabuncu et al, 2010, proposed [56] a nonparametric, probabilistic model for the automatic
segmentation of medical images, given a training set of images and corresponding label maps. Label fusion
segmentation approach can be employed on large multi-subject datasets and yields more accurate segmentation
than FreeSurfer‟s widely used atlas-based segmentation tool and previous label fusion algorithms. It robustly
detected hippocampal volume differences in a study of early Alzheimer‟s Disease and aging.
Debnath Bhattacharyya and Tai-hoon Kim, 2011, proposed [57] an image segmentation method to
indentify or detect tumor from the brain magnetic resonance imaging (MRI) for further consideration of medical
practitioners. Thresholding methods have different result in each image. So a set of image segmentation
algorithms was proposed by which detection of tumor can be done uniquely on brain tumor images.
In contrast to segmentation algorithms, detection algorithms only try to decide if tumor is present and
output the approximate tumor location instead of providing a complete segmentation. The tumor detection could
be used for initializing a segmentation method. Saha et al, 2011, proposed [58] a method to locate the tumor
using a fast unsupervised change detection method searching for dissimilar regions across the symmetry line of
the brain using Bhattacharya coefficient score. This method drew a bounding box, instead of segmenting tumor
which helps in quick analysis of large amounts of data.
Farjam et al, 2012, proposed [59] a template-matching method to detect metastases on conventional
MRI for screening purposes. The result was improved on the spherical template generation process by varying
tumor size, lesion shape and intensities to achieve more accurate detection rates.
The most common way to quantitatively evaluate segmentation results is to calculate the overlap with
the ground truth. Usually, Dice similarity coefficient (DSC) or Hausdroff Distance are used. DSC can range
from 0 to 1 with 0 indicating no overlap and 1 indicating perfect overlap. Another method is to assess results on
a synthetic dataset including ground-truth. Although synthetic data lacks important characteristics of real
images, it has been used by many groups for initially assessing both segmentation and registration methods on
healthy datasets.
Zou et al, 2004, compared [60] the three different validation metrics: area under the receiver operating
characteristics (ROC) curve, mutual information (MI) and Dice similarity coefficient (DSC) for probabilistic
brain tumor segmentation. The conclusion was that for overall classification accuracy the area under the ROC
curve should be used, for sensitivity to changes in tumor size MI was the metric of choice and for spatial
alignment evaluation the Dice coefficient was best.
IV. Conclusion
Image segmentation is extensively used in numerous biomedical-imaging applications, e.g., the
quantification of tissue volumes, study of anatomical structure, diagnosis, localization of pathology, treatment
planning and computer-integrated surgery. Now-e-days, speed of computation is no longer an issue for
researchers. Therefore, the focus is directed toward improvement of information from images obtained through
the slice orientation and perfecting the process of segmentation to get an accurate picture of the brain tumor. As
diagnosing tumor is a complicated and sensitive task; therefore, accuracy and reliability are always assigned
much importance. Hence, an elaborated methodology that highlights new vistas for developing more robust
image segmentation technique is much sought.
References
[1] Cancer Support Community, “Frankly Speaking About Cancer: Brain Tumors”, National Brain Tumor Society, New York, Final
Rep. June 2013.
[2] Charles R. Noback, Norman L. Strominger, Robert J. Demarest and David A. Ruggiero, the Human Nervous System: Structure and
Function, 6th ed., Humana Press, 2005.
[3] The Nervous System, [Online]. Available: http://humananatomy2013.weebly.com/nervous-system.html, Accessed on: April 20,
2015.
[4] 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.
[5] Jan C. Buckner, et al., ―Central Nervous System Tumors, Mayo Clinic Proceedings, Vol. 82, No. 10, pp. 1271-1286. 2007.
[6] Damadian, R., Goldsmith, M. & Minkoff, L. NMR in Cancer: XVI. FONAR Image of the Live Human Body. Physiological
Chemistry and Physics, Vol.9, No.1, pp.97- 100, ISSN: 0031-9325. 1977.
[7] A. O Rodriguez, Principles of Magnetic Resonance Imaging, Revista Mexicana de Fisica, Vol. 50, No. 3, pp. 272-286. 2004.
[8] Rotator Cuff MRI Shoulder Scan & Arthrogram (MRA), [Online]. Available: http://www.colinmcnulty.com/blog/2011/08/04/my-
rotator-cuff-mri-shoulder-scan-and-arthrogram-mra/, Accessed on: April 20, 2015.
[9] [Online]. Available: http://www.radiologyassistant.nl/, Accessed on: April 20, 2015.
[10] [Online]. Available: http://www2.massgeneral.org/livingwithtsc/affects/brain-popup_anatomy.htm, Accessed on: April 20, 2015.
[11] Different types of MRI, [Online]. Available:
https://www.ole.bris.ac.uk/bbcswebdav/institution/Faculty%20of%20Medicine%20and%20Dentistry/MB%20ChB/Radiology/MRI
%20e-tutorial/page_04.htm, Accessed on: April 20, 2015.
[12] Magnetic Resonance Imaging of the Brain, [Online]. Available: http://spinwarp.ucsd.edu/neuroweb/rad401/syllabus.htm Accessed
on: April 20, 2015.
[13] Prima, S., Ayache, N., Barrick, T. & Roberts, N. Maximum Likelihood Estimation of the Bias Field in MR Brain Images:
Investigating Different Modelings of the Imaging Process, Processings of Medical Image Computing and Computer-Assisted
Intervention (MICCAI' 2001), Vol.2208, pp.811-819, DOI: 10.1007/3-540-45468-3_97. 2001.
[14] Li, X., Li, L., Lu, H., Chen, D. & Liang, Z. Inhomogeneity Correction for Magnetic Resonance Images with Fuzzy CMean
Algorithm, Processings of SPIE, Vol.5032, 2003.
[15] Ruan, S., Jaggi, C., Xue, J., Fadili, J. & Bloyet, D. Brain Tissue Classification of Magnetic Resonance Images Using Partial Volume
Modeling. IEEE Transactions on Medical Imaging, Vol.19, No.12, pp.1179- 1187, ISSN: 0278-0062. 2000.
[16] Image Segmentation, [Online]. Available: http://en.wikipedia.org/wiki/Image_segmentation, Accessed on: April 20, 2015.
[17] Sezgin M. and Sankur B., Survey over image thresholding techniques and quantitative performance evaluation, Journal of
Electronic Imaging, Vol. 13, No. 1, pp. 146– 165. 2004.
[18] Rastgarpour M., and Shanbehzadeh J., Application of AI Techniques in Medical Image Segmentation and Novel Categorization of
Available Methods and Tools, Proceedings of the International Multi Conference of Engineers and Computer Scientists 2011 Vol. I,
IMECS 2011, March 16-18, 2011, Hong Kong.
[19] Y. Zhang, H. Qu, Y. Wang, “Adaptive Image Segmentation Based on Fast Thresholding and Image Merging”, Artificial reality and
Telexistence-Workshops, pp. 308-311, 1994.
[20] T. Lindeberg and M.X. Li "Segmentation and classification of edges using minimum description length approximation and
complementary junction cues", Computer Vision and Image Understanding, Vol. 67, No. 1, pp. 88 -98, 1997.
[21] Yian Leng Chang and Xiaobo Li, Adaptive Image Region-Growing, IEEE transaction on image processing, Vol. 3, No. 6,
November, 1994.
[22] W. X. Kang, Q. Q. Yang, R. R. Liang, “The Comparative Research on Image Segmentation Algorithms”, IEEE Conference on
ETCS, pp. 703-707, 2009.
[23] H. Zhang, J. E. Fritts, S. A. Goldman, “Image Segmentation Evaluation: A Survey of unsupervised methods”, computer vision and
image understanding, pp. 260-280, 2008.
[24] H. G. Kaganami, Z. Beij, “Region Based Detection versus Edge Detection”, IEEE Transactions on Intelligent information hiding
and multimedia signal processing, pp. 1217- 1221, 2009.
[25] T.U Paul and S.K. Bandyopadhyay, “Segmentation of Brain Tumor from Brain MRI Images Reintroducing K – Means with
advanced Dual Localization Method”, International Journal of Engineering Research and Applications, June 2012.
[26] G. Heo and C. G. Small. Form representions and means for landmarks: A survey and comparative study. Computer Vision and
Image Understanding, 102(2):188–203, 2006.
[27] 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, pp. 32-57. 1973.
[28] Megha. P. Arakeri · G. Ram Mohana Reddy “Computeraided diagnosis system for tissue characterization of brain tumor on
magnetic resonance images” SpringerVerlag London. 2013.
[29] A. Rajendran and R. Dhanasekaran, A hybrid Method Based on Fuzzy Clustering and Active Contour Using GGVF for Brain
Tumor Segmentation on MRI Images, European Journal of Scientific Research, Vol. 61, No. 2, pp. 305-313. 2011.
[30] Ze-Xuan Ji,Quan-SenSun ,De-ShenXia “A framework with modified fast FCM for brain MR images segmentation” Pattern
Recognition 44, pp. 999–1013. 2011.
[31] Ze-Xuan Ji, Quan-Sen Sun, De-Shen Xia “A modified possibilistic fuzzy c-means clustering algorithm for bias field estimation and
segmentation of brain MR image” Computerized Medical Imaging and Graphics 35, 383–397. 2011.
[32] Pal N.R, Pal K, Keller J.M. and Bezdek J.C, “A Possibilistic Fuzzy c-Means Clustering Algorithm”, IEEE Transactions on Fuzzy
Systems, Vol. 13, No. 4, pp. 517–530, 2005.
[33] Talairach J, Tournoux P. Co-Planar Stereotaxic Atlas of the Human Brain, 3-Dimensional Proportional System: An Approach to
Cerebral Imaging. Stuttgart, Ger. Thieme, p. 122. 1988.
[34] Rajarethinam NC, Andreasen R, Cizadlo T, Arndt S, Swayze VW, et al. Automatic atlas-based volume estimation of human brain
regions from MR images. J. Comput. Assist. Tomogr. 20 pp. 98–106. 1996.
[35] Gibbs P, Buckley D, Blackb S, Horsman A. Tumour determination from MR images by morphological segmentation. Phys Med
Biol, 41, pp. 2437–46. 1999.
[36] Salman Y, Badawi A, Assal M, Alian S. New automatic technique for tracking brain tumor response.Int Conf Biol Med Phys, pp.
1–4. 2005.
[37] Sato M, Lakare S, Wan M, Kaufman A. A gradient magnitude based region growing algorithm for accurate segmentati on. Int Conf
Image Process, 3, pp. 448–51. 2008.