Learning Texture Features from GLCM for Classification of Brain
Tumor MRI Images using Random Forest Classifier
ASHWANI KUMAR AGGARWAL
Department of Electrical and Instrumentation Engineering
Sant Longowal Institute of Engineering and Technology
Longowal-148106, Punjab, INDIA
Abstract—In computer vision, image feature extraction methods are used to extract features so that the features are
learnt for classification tasks. In biomedical images, the choice of a particular feature extractor from a diverse range
of feature extractors is not only subjective but also it is time consuming to choose the optimum parameters for a
particular feature extraction algorithm. In this paper, the focus is on the Grey-level co-occurrence matrix (GLCM)
feature extractor for classification of brain tumor MRI images using random forest classifier. A dataset of brain
MRI images (245 images) consisting of two classes viz. images with tumor (154 images) and images without tumor
(91 images) has been used to assess the performance of GLCM features on random forest classifier in terms of
accuracy, true positive rate, true negative rate, false positive rate, false negative rate derived from the confusion
matrix. The results show that by using optimum parameters, the GLCM feature extracts significant texture
component in brain MRI images for promising accuracy and other performance metrics.
Keywords—statistical texture features, image matching, descriptors, dissimilarity, entropy
Received: April 15, 2021. Revised: February 16, 2022. Accepted: March 20, 2022. Published: April 19, 2022.
1. Introduction
Brain tumors occur due to abnormal growth of cells [1].
Like other tumors, the brain tumors may be benign tumors or
malignant tumors. A malignant tumor usually spreads more
rapidly than a benign tumor [2]. If a brain tumor starts from
the brain, it is known as primary brain tumor, whereas a tumor
which spreads to brain from other parts of the body is known
as secondary brain tumor. The symptoms of brain tumor vary
from a person to person, however, some of the common
symptoms are nausea, vomiting, headache, seizures, and loss
of balance, etc. There are hundreds of types of brain tumors,
the most common types are gliomas, meningioma, and
medulloblastoma [3]. The gliomas brain tumor may be low-
grade astrocytoma which is a slow growing tumor and is
usually benign or glioblastoma multiforme which is rapidly
growing and is usually malignant [4]. The meningioma
usually starts from brain and usually is benign.
Medulloblastoma is more common in children than in adults
[5]. The brain tumor imaging modalities help the radiologists
and oncologists during pre-therapy (for assess the lesion
extent, grading), therapy (delineation), and post-therapy
(therapy response, monitoring) [6]. The brain tumor imaging
modalities include mass spectroscopy, brain perfusion (CT
and MRI), diffusion tensor imaging (DTI), and functional
magnetic resonance imaging (fMRI), etc. [7]. The magnetic
resonance imaging (MRI) uses a very strong magnet and radio
waves to image the brain [8]. All the metal objects need to be
removed before taking MRI scan as the metallic objects
interfere with the magnetic field and causes erroneous signal.
The patients with metallic implants such as pacemaker are
sometimes undergone alternative imaging modality by the
radiologist. The MRI imagers are closed type of open type. In
case of open type MRI scanners, the image quality is not as
good as that obtained with closed type MRI scanners [9]. The
magnet produces a strong magnetic field and rf coils send the
radio waves in the brain. After the radio waves are stopped,
the MRI imager receives the energy signals from the body to
image the brain [10]. The MRI images are sometimes taken
with a dye to enhance the contrast of the image. This paper is
organized as follows. Section II discuss the related work.
Section III presents the dataset and methodology. Section IV
presents the results obtained and the discussion. Section V
concludes the work.
2. Related Work
The automatic detection of tumors in medical images is
challenging task. The feature section and the choice of
classifier is very important for the detection results to give a
significant accuracy. The features derived from the grey-level
co-occurrence matrix have been used to classify the brain
tumors using a two-layered artificial neural network [11]. The
accuracy obtained using the method is 97.5%. The brain
images are classified into four classes viz. astrocytoma,
meningioma, metastatic bronchogenic carcinoma, and
sarcoma. A total of 16 features are extracted based on
correlation, contrast, dissimilarity, energy, homogeneity,
variance, and entropy. The conditional entropy is used using
evolutionary algorithm [12].
The segmentation and the classification of brain tumors
has been done by using texture features derived from GLCM
and using wavelets. The brain MRI images are segmented
using Otsu thresholding method and then k-means clustering
is performed. The texture features are obtained from GLCM
along with features extracted using wavelets. The combined
features are given to PCA before classifying the images with
support vector machines. It is shown that linear kernels
outperform than polynomial kernels for the classification of
brain MRI images [13]. The image retrieval using features
based on ttexture and shape are used [14]. An overview of
methods for brain tumor images using features derived from
grey-level co-occurrence matrix and classified using artificial
WSEAS TRANSACTIONS on SIGNAL PROCESSING
DOI: 10.37394/232014.2022.18.8
Ashwani Kumar Aggarwal
E-ISSN: 2224-3488
60
Volume 18, 2022
neural networks is given [15]. Brain MRI images are classified
using various features obtained from GLCM and kernel
support vector machines as a classifier. The feature reduction
module is used after extraction of features and before giving
the features to the kernel support vector machines for
classification [16]. The classification of brain tumor MRI
images is done using schematic segmentation and support
vector machine-based classification of features extracted from
grey-level co-occurrence matrix [17]. The images are pre-
processed using median filtering. the method is shown to have
accuracy of 93.05% for the classification.
3. Dataset and Methodology
3.1 Dataset
The dataset consists of brain tumor MRI images. The total
number of images is 245. The number of images with tumor
is 154 and the number of images without tumor is 91. The size
of the dataset is 7.22 MB. The image size of images in the
dataset is not same, for example, some of the images have
spatial resolution of 300x168 while others have spatial
resolution of 200x200. The data size is also not same as some
of the images are 4.46 KB while others are 5.76 KB. The
image format of the images in the dataset is jpg format.
3.2 Methodology
The features such as Scale Invariant Feature Transform
(SIFT), Speeded-Up Robust features (SURF), and Oriented
FAST and Rotated BRIEF (ORB) are used for image
classification [18]. These feature extractors extract geometric,
color, textural, and shape information to construct a feature
vector that represents the small neighborhood around the
feature point. SIFT is 128-dimensional feature vector that is
robust to many affine transformations, illumination changes,
and sensor noise [19]. The SIFT features extracted on the brain
tumor MRI images are shown in Fig. 1. As the dimensions of
SIFT are 128 and there are hundreds of such features in a
single image, it becomes computationally expensive to
compute SIFT features for brain tumor MRI image
classification and therefore, SIFT has not been used in this
work.
Fig. 1. Extracted SIFT fesatures
The texture of an image is spatial variation of pixel
intensities in an image. If the pixel intensities vary a lot, that
means the texture is coarse and if the pixel intensities do not
vary much, in that case, the texture is smooth. The texture is
very useful clue for classifying the image contents. The
texture descriptors may be obtained based on the direction,
color, and contrast. The texture analysis of an image can be
done at various resolutions. The texture features can be
classified as statistical, structural, and model based methods.
The statistical texture features are extracted based on
correlation of neighboring pixels, frequency of occurrence of
pairs of pixel intensity values in a neighborhood, and entropy.
The statistical features may be first order features (mean,
median, variance, entropy) or second order features
(relationships between groups of pixels). In this paper, the
texture features are extracted using grey-level co-occurrence
matrix (GLCM). The GLCM is a second order statistical
texture feature extraction method [20]. The GLCM maps the
input image into a table that represents the number of
occurrences of a pair of pixel values at a certain distance and
angle. The angle and distance values are varied over a range
of 0-360 degree and 1 pixel to 8 pixels respectively. The
GLCM is a two-dimensional array where each dimension of
the array is same and is equal to the number of greyscale levels
in the image [21]. For 8-bit image, the number of levels is 256,
and therefore, the GLCM has 256 rows and 256 columns.
There is one such 256x256 matrix for each combination of
distance (offset) and angle value. Each diagonal cell
represents the texture homogeneity in the image, whereas off-
diagonal cells count for texture heterogeneity in the image.
The matrix is made symmetric by adding the matrix with its
transpose. The matrix is then normalized by dividing each
element by the sum of all the elements in the matrix. The
GLCM is used to obtain the second-order statistical texture
features such as angular second moment, entropy, variance,
correlation, inverse difference moment, energy, dissimilarity,
homogeneity, and contrast etc. The energy is obtained by
taking the square root of angular second moment. The
statistical features in an image give several features in the
image that are useful for the classification of brain MRI
images. The following statistical texture features have been
obtained from GLCM for the brain tumor MRI images.
1) Energy feature: The energy feature is the square root
of angular second moment feature. This feature represents
the uniformity in the image texture. The energy feature
obtained from GLCM is given by (1). The maximum value of
the energy is unity as GLCM is a normalized matrix.
2) Dissimilarity Feature: The dissimilarity represents the
heterogeneity in the image texture. The dissimilarity feature
is obtained by multiplying the GLCM cell values with linear
weights as given by (2).
3) Homogeneity Feature: The large homogeneity in the
image texture is represented by large values of the diagonal
elements of the GLCM. If the image pixel values are same,
the homogeneity is maximum. A large contrast reduces the
homogeneity in the image. The homogeneity feature is given
by (3).
.
WSEAS TRANSACTIONS on SIGNAL PROCESSING
DOI: 10.37394/232014.2022.18.8
Ashwani Kumar Aggarwal
E-ISSN: 2224-3488
61
Volume 18, 2022
4) Contrast Feature: The contrast feature is a measure of
difference between the smallest pixel value and the largest
pixel value in a group of pixels. The contrast feature is given
by (4).
A feature vector is obtained from these four features by
extracting these features for various values of offset and
angles. A step size of 1 and 3 pixels is taken and an angle of
0, 45, and 900 degree is taken.
The learnt features are given to a classifier to classify the
brain tumor MRI images. The decision tree, support vector
machines, k-nearest neighbor, and random forest classifiers
are some of the common classifiers used for image
classification tasks. In this paper, we have used random forest
classifier for the classification of the brain tumor MRI images.
The other statistical tests like Chi-square test and t-tests are
conducted to validate the effectiveness of the method.
4. Resuls and Discussion
The dataset is divided into training and testing datasets.
The training and testing sets are prepared from the given
dataset by dividing it into a ratio of 0.8:0.2. The method is
repeated several times to order to make the model robust
towards various types of noise. The features are extracted
from the training dataset. The random forest classifier is
trained using the features extracted from the training dataset.
Thereafter, the features are extracted from testing datasets.
4.1 Actual and predicted Class labels
The classifier is tested for the performance on the features
extracted from the testing dataset. The results of the actual
class labels and predicted class labels for 18 test images is
shown in Fig. 2.
Fig. 2. Actual and Predicted class labels
It is observed that all the images with actual class ‘yes’, are
predicted correctly in all the 9 test images. However, the out
of 9 images with actual class ‘no’, only 6 are predicted
correctly as ‘no’ whereas, rest of the 3 are predicted falsely
as ‘yes’.
4.2 Confusion Matrix
A confusion matrix is obtained for evaluate the
performance of the method. The confusion matrix is put in
Fig. 3.
Fig. 3. Confusion matrix
It is observed from the confusion matrix that there is no
predicted class ‘no’ which is actually ‘yes’. The number of
images tested is 18. Out of 18 images, 9 belong to class ‘yes’
and 9 belong to class ‘no’.
4.3 Performance with various Offset and
Angle of GLCM
The GLCM is obtained for various combinations of steps
(offsets) and the angles to count the number of occurances of
a pair of pixel intensity values. A comparison of calssification
WSEAS TRANSACTIONS on SIGNAL PROCESSING
DOI: 10.37394/232014.2022.18.8
Ashwani Kumar Aggarwal
E-ISSN: 2224-3488
62
Volume 18, 2022
accuracy for various values of steps and angles is given in
Table I.
TABLE I. COMPARISON OF CLASSIFICATION ACCURACY FOR
VARIOUS OFFSET AND ANGLE VALUES OF GLCM
Classification Accuracy
Offset (pixels)
Angle (degree)
32 x 32
64 x 64
1, 3
0, π/4
61.1%
83.3%
1, 3, 5
0, π/4
72.2%
66.7%
1, 3
0, π/4, π/2
72.2%
66.7%
1, 3, 5
0, π/4, π/2
66.7%
72.2%
1, 3, 5, 7
0, π/4
44.4%
66.7%
1, 3, 5, 7
0, π/4, π/2
83.3%
66.7%
The performance of the method drops significantly if the
spatial resolution of the image is reduced. For the same
combination of offset (1 pixel and 3 pixels), and angles (0, π/4),
the accuracy reduces from 83.3% to 61.1% on reducing the
spatial resolution from 64 x 64 to 32 x 32.
5. Conclusion
The brain tumor MRI images are classified into two
classes viz. images with tumor and images without tumor. The
grey-level co-occurrence matrix obtained from the brain
tumor MRI images gives statistical texture features which are
computationally fast as compared to SIFT features. the
random forest classifier classifies the images with an accuracy
of 83.3 % on a limited dataset. the performance of the method
could be improved by applying some pre-processing
operations on the images. Another feature extraction methods
may be combined with GLCM features to enhance the
accuracy. The method can be extended by using additional
layers in the model.
Acknowledgment
The authors are thankful to their colleagues for
proofreading this paper before submission.
References
[1] Surawicz, Tanya S., et al. "Brain tumor survival: results from the
National Cancer Data Base." Journal of neuro-oncology 40.2 (1998):
151-160.
[2] Albini, A., et al. "A rapid in vitro assay for quantitating the invasive
potential of tumor cells." Cancer research 47.12 (1987): 3239-3245.
[3] Waage, Ingunn Syversen, Ingeborg Vreim, and Sverre Helge Torp. "C-
erb B2/HER2 in Human Gliomas, Medulloblastomas, and
Meningiomas: a Minireview." International journal of surgical
pathology 21.6 (2013): 573-582.
[4] A. K. Aggarwal, “GPS-Based Localization of Autonomous Vehicles,”
Autonomous Driving and Advanced Driver-Assistance Systems
(ADAS): Applications, Development, Legal Issues, and Testing, p.
437, 2021.
[5] Mathew, A. Reema, P. Babu Anto, and N. K. Thara. "Brain tumor
segmentation and classification using DWT, Gabour wavelet and
GLCM." 2017 International Conference on Intelligent Computing,
Instrumentation and Control Technologies (ICICICT). IEEE, 2017.
[6] A. Kaur, A. P. S. Chauhan, and A. K. Aggarwal, “An automated slice
sorting technique for multi-slice computed tomography liver cancer
images using convolutional network,” Expert Systems with
Applications, vol. 186, p. 115686, 2021.
[7] Jellinger, K. "Glioblastoma multiforme: morphology and
biology." Acta neurochirurgica 42.1 (1978): 5-32.
[8] Rochkind, Semion, et al. "Extracranial metastases of medulloblastoma
in adults: literature review." Journal of Neurology, Neurosurgery &
Psychiatry 54.1 (1991): 80-86.
[9] A. K. Aggarwal, “Biological Tomato Leaf Disease Classification using
Deep Learning Framework,” International Journal of Biology and
Biomedical Engineering, vol. 16, no. DOI:
10.46300/91011.2022.16.30, pp. 241–244, 2022.
[10] Huang, Raymond Y., et al. "Pitfalls in the neuroimaging of
glioblastoma in the era of antiangiogenic and immuno/targeted
therapy–detecting illusive disease, defining response." Frontiers in
neurology 6 (2015): 33.
[11] Dimou, S., et al. "A systematic review of functional magnetic
resonance imaging and diffusion tensor imaging modalities used in
presurgical planning of brain tumour resection." Neurosurgical
review 36.2 (2013): 205-214.
[12] Katti, Girish, Syeda Arshiya Ara, and Ayesha Shireen. "Magnetic
resonance imaging (MRI)–A review." International journal of dental
clinics 3.1 (2011): 65-70.
[13] M. Garg, J. S. Ubhi, and A. K. Aggarwal, “Neural Style Transfer for
Image within Images and Conditional GANs for Destylization,”
Journal of Visual Communication and Image Representation, 2022.
[14] Foster‐Gareau, Paula, et al. "Imaging single mammalian cells with a
1.5 T clinical MRI scanner." Magnetic Resonance in Medicine: An
Official Journal of the International Society for Magnetic Resonance in
Medicine 49.5 (2003): 968-971.
[15] Armstrong, Peter, and Stephen F. Keevil. "Magnetic resonance
imaging--1: Basic principles of image production." BMJ: British
Medical Journal 303.6793 (1991): 35.
[16] Zulpe, Nitish, and Vrushsen Pawar. "GLCM textural features for brain
tumor classification." International Journal of Computer Science Issues
(IJCSI) 9.3 (2012): 354.
[17] Chauhan, Sumika, Manmohan Singh, and Ashwani Kumar Aggarwal.
"An effective health indicator for bearing using corrected conditional
entropy through diversity-driven multi-parent evolutionary
algorithm." Structural Health Monitoring (2020): 1475921720962419.
[18] Arora, Kratika and Ashwani Kumar Aggarwal. "Approaches for Image
Database Retrieval Based on Color, Texture, and Shape
Features." Handbook of Research on Advanced Concepts in Real-Time
Image and Video Processing, edited by Md. Imtiyaz Anwar, et al., IGI
Global, 2018, pp. 28-50. http://doi:10.4018/978-1-5225-2848-7.ch002
[19] Jain, Shweta. "Brain cancer classification using GLCM based feature
extraction in artificial neural network." International Journal of
Computer Science & Engineering Technology 4.7 (2013): 966-970.
[20] Kadam, Megha, and Avinash Dhole. "Brain tumor detection using
GLCM with the help of KSVM." International Journal of Engineering
and Technical Research 7.2 (2017).
[21] Hussain, Ashfaq, and Ajay Khunteta. "Semantic Segmentation of Brain
Tumor from MRI Images and SVM Classification using GLCM
Features." 2020 Second International Conference on Inventive
Research in Computing Applications (ICIRCA). IEEE, 2020.
WSEAS TRANSACTIONS on SIGNAL PROCESSING
DOI: 10.37394/232014.2022.18.8
Ashwani Kumar Aggarwal
E-ISSN: 2224-3488
63
Volume 18, 2022
Creative Commons Attribution License 4.0
(Attribution 4.0 International, CC BY 4.0)
This article is published under the terms of the Creative
Commons Attribution License 4.0
https://creativecommons.org/licenses/by/4.0/deed.en_US
