Biomedical Research 2018; 29 (16): 3201-3206 ISSN 0970-938X

www.biomedres.info

An intelligent skull stripping algorithm for MRI image sequences using

mathematical morphology.
Kavitha Srinivasan*, Nanditha NM
School of Electrical and Electronics Engineering, Sathyabama University, Chennai, India

Abstract
Brain tumor is a dreadful disease which occurs when abnormal cells form uncontrollably. The modality
adopted to detect abnormalities is Magnetic Resonance Imaging (MRI). MRI brain images contain non-
brain tissues. One of the important preprocessing steps is the whole brain segmentation, the process of
skull stripping which isolates brain tissue and non-brain tissue. Segmentation is tedious and consumes
more time only well experienced radiologist or a clinical expert can perform it with best accuracy. In
order to overcome these limitations, computer aided medical diagnosis is essential. In this work, an
intelligent and a robust skull stripping algorithm using mathematical morphology suited for all types of
MR sequences is proposed. The method was validated on the international database collected from
whole brain Atlas. The performance was evaluated using the metrics Jaccard Similarity Coefficient
(JSC), Dice Similarity Coefficient (DSC), False Positive Rate (FPR), False Negative Rate (FNR),
sensitivity, specificity and accuracy. An average of 97.25% indicates better overlap between proposed
skull stripping and manual stripping by radiologists as a gold standard. The simulation results proved
high accuracy in comparison to the ground truth results which is evident from the similarity coefficient
metrics.

Keywords: Magnetic resonance imaging, Brain segmentation, Morphological operation, Similarity measure.
Accepted on August 21, 2018

Introduction organs, soft tissues, bone and other internal body structure
[4,5].
Brain is the most complex structure of our human body. Grey
matter consists of neuronal cell bodies, neurophil, glial cells, Artifacts introduced during acquisition and undesired tissues
synapses and capillaries which is the major component of the affect the processing quality and lead to inaccurate diagnosis.
central nervous system. White matter is made up of myelinated Thus, an important preprocessing step is skull stripping where
axons called tracts. Grey matter contains numerous cell bodies the brain tissue is segmented from the skull. Since the manual
and has few myelinated axons. Brain is surrounded by tissue segmentation is very time consuming and prone to errors,
membrane called meninges which separates brain from skull. various methods have been developed to automatically remove
CSF is a colorless body fluid that exists in the brain and spinal extra-cerebral tissues without human intervention.
cord. CSF acts as a cushion and provides immunological Morphometric studies require a preprocessing procedure to
protection to the brain [1,2]. isolate the brain from extra-cranial known as skull stripping
[6-9]. Skull stripping ensures better segmentation and assures
Computer aided medical diagnosis with various image accurate diagnosis of brain diseases and probability of
processing techniques has increased the possibility of accurate misclassification of abnormal tissues is also reduced.
prediction due to better visualization. Amongst the various Numerous skull stripping algorithms are available in the
tools, MRI (magnetic resonance image) is an effective imaging literature. Several studies are being carried out on the available
tool, to study the anatomical structure due to its high spatial skull stripping procedures and their performance is analysed on
resolution. Conventional MR sequences are T1-weighted, T2- the standard research datasets. Every algorithm has its own
weighted, and PD-weighted. T1-weighted has highest signal on merits and limitations. In this work, a robust skull stripping
fat tissue and hence better suited for showing anatomical algorithm that is suited for all types of MR image sequences
structures. T2-weighted has highest signal in liquid and hence has been developed.
suited for detecting lesions. PD-weighted is suited for
characterizing lesions by differentiating the contrast of the
Related Works
brain tissues [3]. MRI images brain in axial, coronal and
sagittal planes. MRI scans can produce detailed pictures of From literatures, it is observed that skull stripping can be
performed by morphological operation, intensity based

Biomed Res 2018 Volume 29 Issue 16 3201

 Srinivasan/Nanditha

operation, deformable surface based operation or a

combination of all the methods. Galdame et al. [10] suggested
skull stripping procedure by deformable models and histogram
analysis. Deformable models work on simplex mesh which is
controlled by local gray levels. The performance is compared
with three conventional skull stripping methods such as brain
extraction tool, brain surface extractor, and hybrid watershed
algorithm.
Brummer et al. [11] developed automatic detection of brain
contour with a histogram-based thresholding and conventional
morphological operations with image intensity correction
procedure. Anatomical knowledge is used to discriminate
desired and undesired structures. Gao et al. [12] proposed a
sequence where anisotropic diffusion filtering with edge
detection and morphological operation smoothens noisy
region, identifies the anatomical boundaries and selects the
pixel that correspond to the brain tissue.
Shattuck et al. [13] proposed anisotropic diffusion filtering
combined with edge detection, and mathematical morphology
to remove skull. 2D Marr-Hildreth operator performs edge
detection and low-pass filtering is performed with Gaussian
kernel. Hahn et al. [14] proposed watershed algorithm based
skull stripping which may lead to over-segmentation. Suri [15]
suggested an active contour algorithm where fuzzy
membership function is used to classify brain images. Figure 1. Research database consisting of trans-axial MR brain
Atkins et al. [16] suggested template based method which images of T1-weighted, T2-weighted, PD-weighted and GAD.
separates brain and non-brain regions effectively. It varies with
the type of templates used for separation and also on the Mathematical morphology
application of atlases. Leung et al. [17] performed multi-atlas To analyse geometrical structures such as size, shape,
propagation by combining many segmentations performed by connectivity, mathematical morphology is a tool used which is
atlas registration. based on set theory. It was developed originally for binary
Huang et al. [18] suggested extraction of brain from T1- images but can be used for grayscale and color images. It is a
weighted images through hybridization of expectation collaborative work of Georges et al. at the École des Mines de
maximization algorithm with mathematical morphology and Paris, France. The procedure is to search the image at different
connected component analysis and brain border is found using locations with a pre-defined shape and to decide as how this
geodesic active contours. shape fits or misses the shapes in the image. This is called the
structuring element, and it is also a binary image. Choice of
size and shape of structuring element is need based. Diamond,
Materials and Methods
square, disc, horizontal line, vertical line, cross etc. are the
most commonly used structuring element, erosion and dilation
Patient details and data acquisition are the two essential operations in mathematical morphology.
Research database consists of 27 datasets of 6 patients with age Other operations like opening and closing can be derived from
between 35-71 with anaplastic astrocytoma, grade IV- it.
astrocytoma, low grade glioma, meningioma with T2- and PD-
weighted MR, metastatic carcinoma with T2-weighted MR and Erosion and dilation
grade II- astrocytoma with T1-weighted MR collected from the
Let I be the binary image. S is the structuring element. Erosion
whole brain atlas-Harvard Medical school database (http://
of I by the structuring element S is denoted by IƟS. It denotes
med.harvard.edu/aanlib\) is used to validate the proposed skull
the set of all pixels which S placed at that pixel is contained
stripping algorithm. The RGB image is converted into
within I. Erosion shrinks or performs thinning of the object. It
grayscale before processing. All of the images had 256 × 256
excludes all small unwanted objects in the image.
pixels acquisition matrices, and 2 slices from each patient was
collected. The imaging protocol used a 3 T MRI scanner with IƟS={z ǀ(S) z ⊆ I} → (1)
slice thickness of 1 or 3 mm. The standard slice orientation of
Dilation of I by the structuring element S is denoted by I ⊕ S. I
trans-axial is considered. Figure 1 represents the complete set
overlaps by at least one element. Dilation performs thickening
of database acquired.
of an image. It highlights small objects in the image.

3202 Biomed Res 2018 Volume 29 Issue 16

An intelligent skull stripping algorithm for MRI image sequences using mathematical morphology

I ⊕ S=*z ǀ( ̂) z ∩ I ≠ ϕ} → (2) information is eliminated and used to focus on the area of

interest [22-24]. Structuring Element (SE) is so chosen to have
Opening and closing size and shape according to the type of MR image sequence.
For each pixel location, morphological operation is performed
Opening operation denoted by I • S smoothens the contour of between SE and MR image. Figure 2 shows the flowchart of
an object and breaks the narrow lines. In opening operation the algorithm using morphological operations.
erosion is performed followed by dilation
Step 1: Conversion to binary image: The input image is
I • S=(A Ɵ S) ⊕ S → (3) converted to binary image with Otsu’s thresholding [24]. Otsu's
Closing operation denoted by I • S smoothens the contours. It method automatically performs reduction of a gray level image
eliminates discontinuity and small gaps between the objects. In to a binary image. The algorithm considers the image as two
closing operation, dilation is performed followed by erosion. classes of pixels, foreground pixels and background pixels.
Optimum threshold is obtained which separates the two classes
I • S=(A ⊕S) Ɵ S → (4) so that their intra-class variance is minimal and their inter-class
variance is maximal.
Step 2: Creation of mask: Brain is the largest connected
component. This component is extracted from the binary
image. Dilation and erosion operations are performed to
preserve the minute features of the brain in the resultant image.
By filling the holes, the brain becomes a complete connected
component.
Step 3: Superimposition: The final skull stripped image is
obtained by superimposing the mask on the input image.

Results and Discussion

Skull stripping is a challenging pre-processing step in tumor
characterization due to intrinsically imprecise nature of brain
image. It helps to remove the skull from the input image. The
performance of the proposed approach is compared with
ground truth which is a manual stripping by a well experienced
radiologist considered as gold standard. Figure 3 shows the
results of the proposed skull stripping algorithm for different
MR brain image sequences. The algorithm is carried out by
using Matlab tool kit.
Seven performance metrics namely, Jaccard similarity
coefficient (JSC), Dice similarity coefficient (DSC), false
positive rate (FPR), false negative rate (FNR), sensitivity,
specificity and segmentation accuracy have been calculated.
JSC and DSC measure the similarity between ground truth and
skull stripped images. The confusion matrix in terms of TP,
TN, FP, and FN between the expected output and ground truth
are shown in Table 1. TP represents the number of pixels that
are correctly classified as brain tissue; TN is the number of
pixels that are correctly classified as non-brain tissue; FP is the
number of pixels that are in-correctly classified as brain tissue
and FN is the number of pixels that are in-correctly classified
as non-brain tissue [25-27], JSC can be formulated according
to the following equation
Jaccard coefficient JSC (A, B)=(A ∩ B)/(A ∪ B) → (5)
Figure 2. Flow chart of skull stripping algorithm using
=2TP/2TP+FP+FN
morphological operations.
where A is the area of the brain region in the ground truth
Proposed skull stripping algorithm skull-stripped image and B is the area of the brain region of the
corresponding image with the proposed skull stripping
Proposed skull stripping algorithm works on morphological
operations [19-21]. With morphological operation, extra

Biomed Res 2018 Volume 29 Issue 16 3203

 Srinivasan/Nanditha

algorithm. If JSC is 1, it represents complete overlap, whereas

an index of 0 represents that there are no overlapping pixels.
DSC is used to describe the overall level of similarity between
proposed skull-stripped algorithm and ground truth. DSC has
been calculated according to the following equation
Dice coefficient DSC (A, B)=2 (A ∩ B)/|A|∪|B| → (6)
=TP/TP+FP+FN
False positive rate (FPR) and false negative rate (FNR) were
used to quantify over and under segmentation. Both FPR and
FNR were calculated according to
False positive rate FPR (A, B)=(A/B)/(A ∪ B) → (7)
=FP/TN+FP
False negative rate FNR (A, B) = (B/A)/(A ∪ B) → (8)
=FN/TP+FN
A direct relation between JSC, FPR, and FNR is according to
the following expression:
Jaccard coefficient JSC (A, B)=1-FPR-FNR → (9)
Sensitivity=FP/TN+FP → (10)
Specificity=TN/TN+FP → (11)
Accuracy=TP+TN/TP+TN+FP+FN → (12) Figure 3. Results of the proposed skull stripping algorithm for
different MR brain image sequences.
The performance metrics evaluated for the proposed skull
stripping algorithm is given in Table 2. The proposed method Table 1. Confusion matrix defining the terms TP, TN, FP, FN.
is robust and hence successfully strips the skull for all types of
brain image sequences. Average performance using dice Expected outcome Ground truth Row total

coefficient is 97.25% and using Jaccard coefficient is 94.87% Positive Negative

which illustrates that highest similarity or overlap between the
Positive TP FP TP+FP
skull stripping algorithm proposed and the ground truth.
Negative FN TN FN+TN

Column total TP+FN FP+TN TP+FP+FN+TN

Table 2. Performance of skull stripping algorithm.

Subject Jaccard coefficient Dice coefficient Similarity Specificity Accuracy

T1_ weighted 0.8012 0.8896 0.8059 1 0.9995

T1_ weighted 0.8585 0.9239 0.9502 0.9995 0.9993

T2_ weighted 0.9482 0.9734 0.973 0.9981 0.9964

T2_ weighted 0.9982 0.9991 0.999 0.9999 0.9998

T2_ weighted 0.9592 0.9792 0.9896 0.9975 0.997

T2_ weighted 0.9964 0.9982 0.9992 0.9998 0.9997

T2_ weighted 0.9987 0.9993 0.9992 0.9999 0.9998

T2_ weighted 0.9986 0.9993 0.9992 0.9999 0.9998

T2_ weighted 0.9649 0.9822 0.9653 0.9999 0.9922

T2_ weighted 0.9752 0.9875 0.9756 0.9999 0.9945

T2_weighted 0.9953 0.9976 0.9988 0.9991 0.9991

3204 Biomed Res 2018 Volume 29 Issue 16

An intelligent skull stripping algorithm for MRI image sequences using mathematical morphology

PD-weighted 0.9804 0.9901 0.9812 0.9998 0.9963

PD-weighted 0.9871 0.9935 0.9873 0.9999 0.9972

PD-weighted 0.9772 0.9885 0.9772 1 0.9989

PD-weighted 0.9869 0.9934 0.9931 0.9998 0.9996

PD-weighted 0.9143 0.9552 0.9154 0.9968 0.9918

GAD 0.9616 0.9804 0.9641 0.9997 0.9962

GAD 0.9658 0.9826 0.9684 0.9997 0.9971

GAD 0.8058 0.8925 0.9453 0.9967 0.9957

GAD 0.9293 0.9634 0.9962 0.9991 0.9991

GAD 0.7454 0.8541 0.7788 0.9998 0.9987

GAD 0.9813 0.9905 0.9905 0.9998 0.9996

GAD 0.9894 0.9942 0.9902 0.9997 0.9983

GAD 0.9799 0.9899 0.9807 0.9999 0.9972

GAD 0.9683 0.9839 0.9687 0.9999 0.9929

GAD 0.976 0.9879 0.9769 0.9997 0.9946

GAD 0.9722 0.9859 0.9797 0.9968 0.9918

Average performance 0.9487 0.9725 0.9648 0.9992 0.9971

Conclusion automated skull-stripping methods applied to

contemporary and legacy images: effects of diagnosis,
Skull stripping being a preliminary step, its accuracy is a key bias correction and slice location. Hum Brain Mapp 2006;
factor. Most of the methods suggested in literature cannot be 27: 99-113.
applied for all brain image sequences and orientations. In this
7. Matsumoto S, Asato R, Konishi J. A fast way to visualize
paper, an intelligent skull stripping algorithm that works well
the brain surface with volume rendering of MRI data. J
for all types of MR Image sequences such as T1, T2, PD, and
Digit Imaging 1999; 12: 185-190.
GAD is proposed. Experimental results confirm the efficacy of
8. Mahmood Q, Chodorowski A, Mehnert A, Gellermann J,
the proposed algorithm. Moreover comparison with ground
Persson M. Unsupervised Segmentation of Head Tissues
truth showed that the proposed approach is promising in
from Multi-modal MR Images for EEG Source
obtaining high performance metrics.
Localization. J Digit Imaging 2015; 28: 499-514.
9. Hata Y, Kobashi S, Kondo K,Kitamura YT, Yanagida T.
References
Transcranial ultrasonography system for visualizing skull
1. Purves D, Augustine GJ, Fitzpatrick D, Hall WC, and brain surface aided by fuzzy expert system. IEEE
Lamantia AS, McNamara JO, White LE. Neuroscience Trans Syst Man Cybern 2005; 35: 1360-1373.
(4th Edn.). Sinauer Assoc 2008; 15-16. 10. Benson CC, Lajish VL, Kumar R. A novel skull stripping
2. Blumenfeld H. Neuroanatomy through clinical cases (2nd enhacement algorithm for the improved brain tumor
Edn.). Sunderland, Mass.: Sinauer Assoc 2010; 21. segmentation using Mathematical morphology. Int J
3. Haacke EM, Brown RW, Thompson MR, Venkatesan R. Image Graph Sig Proc 2016; 7: 59-66.
Magnetic resonance imaging, physical principles and 11. Ahmad C, Camel T. Quantitative evaluation of robust
sequence design. John Willey Sons NY 1999. skull stripping and tumor detection applied to axial MR
4. Quencer RM, Bradley WG. MR imaging of the brain: images. Brain Inform 2016; 3: 53-61.
what constitutes the minimum acceptable capability? 12. Otsus N. A threshold selection method from gray level
AJNR Am J Neuroradiol 2001; 22: 1449-1450. histograms? IEEE Transac Sys MAN Cybern 1979; 9.
5. Cheour M. Advantages of brain MRI. Radiology Org 13. Torheim T, Malinen E, Kvaal K, Lyng H, Indahl UG,
2010. Anderson EKF, Futsaether CM. Classification of dynamic
6. Fennema-Notestine C, Ozyurt IB, Clark CP, Morris S, contrast enhanced MR images of cervical cancers using
Bischoff-Grethe A, Bondi MW, Jernigan TL, Fischl B, texture analysis and support vector machines. IEEE Trans
Segonne F, Shattuck DW, Leahy RM, Rex DE, Toga AW, Med Imaging 2014; 33: 1648-1656.
Zou KH, Brain M, Brown GG. Quantitative evaluation of

Biomed Res 2018 Volume 29 Issue 16 3205

 Srinivasan/Nanditha

14. Dhanalakshmi K, Rajamani V. An intelligent mining 23. Razmjooy N, Mehdi R, Noradin G. Imperialist
system for diagnosing medical images using combined competitive algorithm-based optimization of neuro-fuzzy
texture histogram features. Int J Imaging Sys Technol system parameters for automatic red-eye removal. Int J
2013; 23: 194-203. Fuzzy Sys 2017; 19: 1144-1156.
15. Rajendran P, Madheswaran M. Pruned associative 24. Razmjooy N, Sheykhahmad FR, Ghadimi N. A hybrid
classification technique for the medical image diagnosis neural network-world cup optimization algorithm for
system. Proc 2nd Int Conf Mac Vis 2009; 293-297. melanoma detection. Op Med 2018; 13: 9-16.
16. Kalavathi P, Prasath VB. Methods on skull stripping of 25. Atkins MS, Siu K, Law B, Orchard JJ, Rosenbaum WL.
MRI head scan images-a review. J Digit Imaging 2016; Difficulties of T1 brain MRI segmentation techniques,
29: 365-379. medical imaging. Proc SPIE 2001; 4684: 1837-1844.
17. Brummer ME, Mersereau RM, Eisner RL, Lewine RJ. 26. Leung KK, Barnes J, Modat M, Ridgway GR, Bartlett JW,
Automatic detection of brain contours in MRI data sets. Fox NC, Ourselin S. Brain MAPS: an automated, accurate
IEEE Trans Med Imaging 1993; 12: 153-166. and robust brain extraction technique using a template
18. Gao J, Xie M. Skull stripping MR brain images using library. Neuroimage 2011, 55: 1091-1108.
anisotropic diffusion filtering and morphological 27. Huang A, Abugharbieh R, Tam R, Traboulsee A. MRI
processing. Proc Int Sym Comp Netw Multimed Technol brain extraction with combined expectation maximization
Wuhan 2009; 1: 1-4. and geodesic active contours. Proc. IEEE Int Symp Signal
19. Shattuck DW, Sandor-Leahy SR, Schaper KA, Rottenberg Proc Inf Technol 2006; 107: 107-111.
DA, Leahy RM. Magnetic resonance image tissue
classification using a partial volume model. Neuroimage *Correspondence to
2001; 13: 856-876.
Kavitha Srinivasan
20. Hahn HK, Peitgen HO. The skull stripping problem in
MRI solved by single 3D watershed transform. Proc Med School of Electrical and Electronics Engineering
Image Comput Comp Assist Interv 2000; 134-143.
Sathyabama University
21. Suri JS. Two-dimensional fast magnetic resonance brain
segmentation. IEEE Eng Med Biol Mag 2001; 20: 84-95. India
22. Ali P, Ramezani M, Ghadimi N. A hybrid neural network-
gray wolf optimization algorithm for melanoma detection.
Biomed Res 2017; 28: 8.

3206 Biomed Res 2018 Volume 29 Issue 16