ArticlePDF Available

GLCM texture analysis on different color space for pterygium grading

Authors:
M.Z.C. Azemin
M.Z.C. Azemin
M.Z.C. Azemin
  • This person is not on ResearchGate, or hasn't claimed this research yet.
Mohd Izzuddin Mohd Tamrin at International Islamic University Malaysia
Mohd Izzuddin Mohd Tamrin
Mohd Radzi Hilmi at International Islamic University Malaysia
Mohd Radzi Hilmi
Khairidzan Kamal at International Islamic University Malaysia
Khairidzan Kamal
Download full-text PDF
Read full-text
Download citation

Abstract and Figures

GLCM texture features have been widely used to characterize biomedical images. Most of the previous studies using GLCM features to characterize biomedical images only consider single or limited color space due to the use of only one color model. To mimic human color perception, conventional RGB color model may need to be supplemented with other color space models for better human vision representation. This study is aimed to find an optimal set of GLCM features extracted from different color space for pterygium grading. Mimicking human color perception has commonly employed RGB color space, which is shown in this paper is inadequate. GLCM features when extracted in various color space show better representation of human perception (correlation coefficient > 0.6) compared to using RGB color space (correlation coefficient < 0.2).
Figures - uploaded by Mohd Radzi Hilmi
Author content
All figure content in this area was uploaded by Mohd Radzi Hilmi
Content may be subject to copyright.
Content uploaded by Mohd Radzi Hilmi
Author content
All content in this area was uploaded by Mohd Radzi Hilmi on Apr 10, 2019
Content may be subject to copyright.
VOL. 10, NO. 15, AUGUST 2015 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.
www.arpnjournals.com
6410
GLCM TEXTURE ANALYSIS ON DIFFERENT COLOR SPACE FOR
PTERYGIUM GRADING
Mohd Zulfaezal Che Azemin1, Mohd Izzuddin Mohd Tamrin2, Mohd Radzi Hilmi1 and
Khairidzan Mohd Kamal3
1Kulliyyah of Allied Health Sciences, International Islamic University Malaysia, Kuantan, Malaysia
2Kulliyyah of ICT, International Islamic University Malaysia, Gombak, Malaysia
3 Kulliyyah of Medicine, International Islamic University Malaysia, Kuantan, Malaysia
ABSTRACT
GLCM texture features have been widely used to characterize biomedical images. Most of the previous studies
using GLCM features to characterize biomedical images only consider single or limited color space due to the use of only
one color model. To mimic human color perception, conventional RGB color model may need to be supplemented with
other color space models for better human vision representation. This study is aimed to find an optimal set of GLCM
features extracted from different color space for pterygium grading. Mimicking human color perception has commonly
employed RGB color space, which is shown in this paper is inadequate. GLCM features when extracted in various color
space show better representation of human perception (correlation coefficient > 0.6) compared to using RGB color space
(correlation coefficient < 0.2).
Keywords: GLCM texture, biomedical, pterygium grading.
INTRODUCTION
A measurable system for looking at pixel
composition that regards the spatial associations of pixels
is known as Gray-Level Co-occurrence Matrix (GLCM)
[1]. The matrix representation operates as textural
descriptor by ascertaining how frequently a pixel sequence
matches with a particular pattern exists in a two-
dimensional image, generating a GLCM. Features based
on statistical formula can further be applied on GLCM to
quantify the texture pattern of the image, for example,
correlation and homogeneity analyses.
GLCM texture features have been widely used to
characterize biomedical images. Previous researches
include classification of benign and malignant liver tumors
[2], lung image registration [3], diagnosis in breast MRI
images [4], identification of lobar fissure regions [5], and
in the classification of epithelial pre-cancer cells [6].
Most of the previous studies using GLCM
features to characterize biomedical images only consider
single or limited color space due to the use of only one
color model. To mimic human color perception,
conventional RGB color model may need to be
supplemented with other color space models for better
human vision representation. Previous studies have shown
that different color space is more perceptually relevant
depending upon its applications [7]. Prior research works
using different color space mostly focus on solving image
segmentation problems [8–10].
Tissue redness has been an important indicator in
assessing and diagnosing disease [11–13]. The importance
of automating the redness grading has been highlighted
previously particularly due to low intra-graders
repeatability [14]. However, there is a lack of research to
employ GLCM features on different color space to mimic
human clinical grading. This study is aimed to fill the gap
in the literature.
MATERIALS AND METHODS
A total of 68 eye images affected by pterygium
are captured with slit lamp bio microscopy using a
diffused white light. Figure-1 shows the images with their
respective grades labelled by a clinician. The region of
interest is from the apex of the pterygium to the limbus as
shown in the Figure-1(a). The grading is treated as a
continuous variable.
Figure-1. Pterygium grading based on the tissue redness.
(a) Grade 1.0, (b) Grade 1.5, (c) Grade 2.0, (d) Grade 2.5,
(e) Grade 2.5, (f) Grade 3.0.
VOL. 10, NO. 15, AUGUST 2015 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.
www.arpnjournals.com
6411
The previous research [14] used RGB colour
space model as their feature for texture analysis. In this
paper, 14 different colour space models are employed as
the features for evaluating the quality of pterygium
grading. The full list of the colour spaces is available from
Table-1.The motivation of using these colour space
models is to mimic the real perception of human expert in
evaluating these images for grading the pterygium.
Table-1. List of color space used in this work.
The steps undertaken to extract these features using these
colour spaces are summarized in Figure-2. For every
image and its selected region of interest that are marked
for pterygium grading are retrieved in Red, Green and
Blue colours. Next, the Matlab Toolbox developed by
Pascal [15] is used for converting this image from the
RGB colour model to the other 14 colour space models. A
matrix is created to split every colour spaces into its 3
components. The total number of textures for the entire
colour space models that are produced for each image is
168. Following this, only the selected region of interest
are considered for the conversion to GLCM and the
remaining pixels residing outside these area are assigned
as not a number (NaN).
Figure-2. Flow chart on extracting features based on
GLCM properties.
Figure-3 demonstrates the conversion to GLCM.
The value of each pixel is set between the ranges of 1 to 8,
which indicates the level of intensity for each colour
component in the image. The spatial relationship between
each pixel from the region of interest is determined by
calculating the frequency of the intensity of that pixel in i
against its adjacent pixel in j.
Figure-3. The construction of GLCM (right) from a
hypothetical image with varying intensity (left).
VOL. 10, NO. 15, AUGUST 2015 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.
www.arpnjournals.com
6412
For example, in Figure-3, the value 2 in element
(7,7) from the GLCM is the total sum of co-occurrence
between the adjacent pixels with the same intensity of 7
throughout the image. The algorithm will scan from these
pixels from left to right until it reaches the end of the
pixel. The remaining elements from the GLCM have the
value of 1 and 0 because the sum of co-occurrence
between its respective adjacent pixels is 1 or 0.
The GLCM produced is used to statistically
examine spatial distribution of the pixels’ intensity levels
in this image. The statistical analyses employed for the
texture analysis are contrast, correlation, energy,
homogeneity. The details on these statistical properties
are described below.
Contrast measures the differences in intensity
between a pixel against the remaining pixels in the image.
It is defined as,
Contrast =
where p(i,j) is the position of the GLCM in that the value
represent the sum of co-occurrence between adjacent
pixels of i and its neighbour j.
Correlation measure the level of correlations
between a pixel against the remaining pixels in the image.
It is defined as,
Correlation =
Energy measure the summation of squared
element in the entire GLCM. It is formulated as,
Energy = .
Homogeneity measure the similarity in the
variation of the distribution in the GLCM against the
diagonal of the matrix. It is defined as,
Homogeneity = .
RESULTS
Figure-4 shows the results of correlation analysis
between the GLCM features and scores from the human
grading. Four features are identified as good features with
correlation coefficients more than 0.6. The features are
GLCM-Contrast (Pr component of Y’PbPr color space),
GLCM-Energy (Pr component of Y’PbPr color space),
GLCM-Homogeneity (Pr component of Y’PbPr color
space), and GLCM-Energy (V component of Y’UV color
space). The features extracted from conventional RGB
color space perform poorly with correlation coefficients
less than 0.2.
Figure-4. Correlation between human grading and the
GLCM features extracted from images with different color
space.
CONCLUSIONS
This research work identifies a set of good
GLCM features that can be used in the future on modeling
human perception of tissue redness. It has been proven
that GLCM features extracted from conventional RGB
color space are not sufficient and must be supplemented
with features in other color space. GLCM features when
extracted in various color space show better representation
of human perception (correlation coefficient > 0.6)
compared to using RGB color space (correlation
coefficient < 0.2).
However, with a single feature, the best GLCM
feature can only account for 49.63% (from r-squared
analysis) variability of the human grading. Fusion of
multiple features is warranted to produce a better model of
pterygium grading. Experiments with different type of
textural features which include fractal dimension [16, 17]
may also further be tested in future research.
REFERENCES
[1] Baraldi A. and Parmiggiani F. 1995. Investigation of
the textural characteristics associated with gray level
cooccurrence matrix statistical parameters. IEEE
Trans. Geosci. Remote Sens. Vol. 33, 293–304.
[2] Xian G.M. 2010. An identification method of
malignant and benign liver tumors from
ultrasonography based on GLCM texture features and
fuzzy SVM. Expert Syst. Appl. Vol. 37, pp. 6737–
6741.
[3] Park S., Kim B., Lee J., Goo J.M. and Shin Y.G.
2011. GGO nodule volume-preserving nonrigid lung
registration using GLCM texture analysis. IEEE
Trans. Biomed. Eng. Vol. 58, pp. 2885–2894.
[4] Yachun P., Yuanzhi S. and Li L. 2010. Breast lesion
classification on MRI by texture features. 2nd
VOL. 10, NO. 15, AUGUST 2015 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved.
www.arpnjournals.com
6413
International Conference on Information Science and
Engineering, ICISE2010 - Proceedings.
[5] Wei Q. and Hu Y. 2009. A study on using texture
analysis methods for identifying lobar fissure regions
in isotropic CT images. Conf. Proc. IEEE Eng. Med.
Biol. Soc. pp. 3537–3540.
[6] Zheng W., Li D., Li S., Zeng Y., Yang Y. and Qu J.Y.
2011. Diagnostic value of nonlinear optical signals
from collagen matrix in the detection of epithelial pre
cancer.
... However, studies on the morphological features of pterygium, such as redness, which could also act as an indicator of disease severity are lacking. Pterygium Redness Grading Software (PRGS) was developed to address this issue by introducing redness as an indicator to determine the effects of different types of pterygia on corneal curvature Che Azemin et al. 2015Mohd Radzi et al. 2017). ...
... Preliminary studies Che Azemin et al. 2015 attempted to quantify and describe pterygium redness using extensive image analysis combining colour-space conversion, colour layers, textural descriptors, and feature selection technique. The introduction of the semi-automated PRGS has allowed the grading of pterygium redness on a continuous scale of 1 (minimum redness) to 3 (maximum redness) (Mohd Radzi et al. 2017). ...
... We used digitised pterygium images and other data obtained during the baseline visit. image acquisition, analyses, and clinical grading were performed as previously described (Che Azemin et al. 2015Azemin et al. , 2014hilmi et al. 2019;Mohd Radzi et al. 2017). All participants were selected on the basis of the following specific criteria: participants with an established diagnosis of primary pterygium; participants of both sexes with ages ranging from 20 to 70 years; and those without any history of ocular trauma, ocular surgery, contact lens wear, and any ocular anterior segment disease other than pterygium which may affect vision (Che Azemin et al. 2015Azemin et al. , 2014hilmi et al. 2019;Mohd Radzi et al. 2017). ...
View
... Random forest classifier [34] It is in ensemble and it can be processed by using the RandomForestClassifier() function. ...
View
... Comparative analysis between formalin-fixed and fresh pterygium tissue via freeze-dried method based on independent T-test of angiogenesis activity (Qi et al., 2012;Sato et al., 2014;Che Azemin et al., 2014;Azemin et al., 2015;Azemin et al., 2016;Kumagai et al., 2016), which served as an important indicator for fibrous tissue growth, which in this case leads to various degree of induced corneal astigmatism. This may explain why some pterygium cause significant corneal astigmatism while others do not. ...
Measurement of Pterygium Tissue Dry Weight Using Two Different Tissue Preparation Techniques in Freeze-Dry Method
Article
Full-text available
  • Dec 2022
Due to pterygium tissue compression on the corneal surface, it has been hypothesised that the degree of corneal astigmatism caused by pterygium can vary. This study aims to develop a freeze-dried method for estimating net pterygium tissue mass (NPTM) as dry weight. A single surgeon (KMK) excised 60 primary pterygium using the controlled partial avulsion technique and divided them into two groups: formalin-fixed (n=30) and unfixed (n=30). After determining the weight of each sterile container, 5 mL of 5% buffered formaldehyde was added to the formalin-fixed group and stored for one week, while 5 mL of distilled water was added to the non-fixed pterygium group. Each container was pre-frozen for 12 hours before being freeze-dried (-24 hours at -80 degrees Celsius). The result is referred to as the net pterygium tissue mass (NPTM). Using an independent T-test, a comparison of wet and dry weight and percentage of NPTM was conducted between groups. Wet weights for formalin-fixed and non-fixed pterygium were 253.33 82.17 g and 255.17 63.52 g, respectively, while dry weights were 184.92 84.31 g and 179.54 72.85 g. Formalin-fixed pterygium tissue revealed a slightly higher percentage of NPTM than non-fixed pterygium tissue (69.39 13.29% vs. 67.75 13.29%, p = 0.792), but this difference was not statistically significant. The freeze-dried method can be utilised to quantify the NPTM of pterygium fibrovascular tissue and investigate the influence of pterygium translucency on predicting induced-corneal astigmatism.
View
... All participants in this study were selected based on inclusion criteria including an established diagnosis of primary pterygium, patients of either gender with ages between 20 and 70 years, and no history of ocular trauma, ocular surgery, and contact lens wear [1,[14][15][16]. Patients with significant ocular surface diseases, such as recurrent pterygium, corneal irregularity, and opacity due to diseases other than pterygium, were excluded [13]. ...
View
... All participants in this study were selected based on inclusion criteria including an established diagnosis of primary pterygium, patients of either gender with ages between 20 to 70 years, and no history of ocular trauma, ocular surgery and contact lens wear [1,[14][15][16]. Patients with significant ocular surface diseases, such as recurrent pterygium, corneal irregularity, and opacity due to diseases other than pterygiumwere excluded [13]. ...
View
... The grey-level co-occurrence matrix consists of elements describing the frequency of pairs of pixels of a given brightness and neighbourhood. Based on this matrix energy, contrast, correlation, summation average, summation and differential variance, and entropy and differential entropy are calculated [5,[10][11][12]23,38]. ...
Analysis of X-rays in bone remodelling around Active and Replace dental implants
Article
Full-text available
  • Feb 2022
Purpose of the study: The purpose of the study was to examine on the basis of radiographic images of bone tissue, whether there are differences in the rate of bone remodelling using different shapes of implants in the mandible and maxilla. Moreover, the study also compares texture features obtained on the basis of these images for healthy bone tissue, bone directly after implantation and after a 12-month period of prosthetic loading. Materials and Methods: The subject of the analysis was radiovisiogram images obtained from the Medical University of Bialystok from the Department of Dental Surgery. They are radiovisiogram photographs of 146 people aged 18–74, treated implantally due to missing teeth. The whole group of patients received two types of implants (Active and Replace) of the same company, made of titanium, intraosseous, screw-in. Results: It has been shown that both in the upper jaw and the mandible, the values of texture parameters obtained for bone images made after one year of prosthetic loading are closer to healthy tissue than immediately after implantation. These values for the mandible were relatively closer to those obtained on the basis of healthy tissue than those for the upper jaw. The bone around the implant with a single threading achieved better results in the mandible than the one with a double threading. Conclusion: The type of bone tissue and the shape of the implant have an impact on the achieved osseointegration. With the passage of time and the process of bone remodelling, the damaged tissue returns to its normal structure.
View
... The inclusion criteria were an established diagnosis of primary pterygium, patients of either gender with ages between 20 and 70 years, and no history of ocular trauma, ocular surgery, and contact lens wear. [21][22][23][24] Patients with significant ocular surface diseases, such as recurrent pterygium, corneal irregularity, and opacity due to diseases other than pterygium, were excluded. Patients for whom corneal topography could not provide reproducible measurements due to obstruction of the central cornea by pterygium were also excluded. ...
Pterygium recurrence and corneal stabilization point after pterygium excision using the controlled partial avulsion fibrin glue technique
Article
Full-text available
  • May 2020
Background: This study aimed to evaluate the pterygium recurrence rate and corneal stabilization point after pterygium excision via the controlled partial avulsion fibrin glue technique using multiple corneal parameters. Methods: One hundred eyes of 100 patients who had undergone primary pterygium excision surgery via the controlled partial avulsion fibrin glue technique were retrospectively reviewed. Corneal stabilization points were determined over four follow-up sessions (i.e., the 1st, 3rd, 6th, and 12th months after surgery) based on changes in Simulated-K, corneal irregularity measurement, shape factor, and toric mean keratometry. Post-operative courses were followed for 12 months after surgery. Recurrence was defined as the regrowth of fibrovascular tissue 1 mm past the corneoscleral limbus. Results: No sign of pterygium recurrence and the corneal stabilization point were observed at the third month post-operation. Significance improvements in all corneal parameters were noted between the 1st and 3rd months (both p < 0.001); however, insignificant changes were noted at the following 6th- and 12th-month visits (both p > 0.05). Conclusion: The controlled partial avulsion fibrin glue technique may improve surgical outcomes with long-term recurrence rates equal to or lower than those previously reported. Corneal surface recovery is completed after the third month of the excision procedure.
View
Repeatability and Reproducibility of Pterygium Grading Based on Clinical Translucence Appearance
Article
Full-text available
  • Oct 2021
Background: The goal of this study was to evaluate the repeatability and reproducibility of pterygium grading based on clinical translucence appearances and its reliability in clinical setting. Methods: A total of 93 primary pterygium eyes from 93 participants who visited a tertiary ophthalmic clinic were recruited in this study. Two (2) ophthalmologists and two (2) optometrists evaluated and graded the 93 primary pterygium images in randomized fashion. Graders were instructed to utilise the clinical translucence appearance of pterygium to grade them into type I, II and III. Repeatability testing was done by a single expert by comparing grading of each image on two separate sessions, with one month interval between sessions. Reproducibility was tested by comparing the grading obtained by both experts and optometrists. Results: Paired and independent T-test results showed no significance difference between graders for both experts and optometrists’ group (all P > 0.05). Intra-grader and inter-grader reliability estimates were high with intraclass correlation ranging from 0.97 to 0.98 for both groups. Conclusions: Pterygium clinical grading based on its translucence appearance is reliable and repeatable in clinical setting, easily to be graded, interpreted, and recommended for clinicians with different levels of experience.
View
Reliability analysis of subjectively graded real-image pterygium based on its translucence appearance between young and experi-enced clinicians
Article
Full-text available
  • Oct 2021
Introduction: Pterygium is commonly subjectively evaluated via anterior segment assessments during slit-lamp examination. Thus, this assessment prones individual variations as it requires subjective grading and adequate experience to ensure consistency of diagnosis and management. Purpose: This study aimed to compare the reliability of subjectively graded real-image pterygium based on its translucence appearance between experienced clinicians. Design of study: Prospective randomized study. Materials and methods: Thirty (30) primary pterygium images from 30 pterygium patients were captured in a standardized magnification, illumination and formatting setting as previously de-scribed. All images were projected using PowerPoint presentation™ on a liquid crystal display (LCD) monitor with standard resolution. Two clinicians with different levels of experiences act as a grader and grade all images based on reference images provided. For reliability testing, intra-grader assessment was repeated twice with different sequences at least a month apart between each session. Both clinicians were given a set of 30 randomized pterygium images for all sessions. Reliability testing was evaluated using paired T-test and independent T-test. Results: Descriptive analysis revealed observer 1 obtained mean grade of 2.19 (SD = 0.670) and 2.23 (SD = 0.713) for session 1 and 2 respectively. Observer 2 obtained 2.04 (SD = 0.853) and 2.08 (SD = 0.894) for session 1 and 2 respectively. Paired T-test showed the difference for both observers in both sessions were not statistically significant (P = 0.776 and P = 0.583) respectively. Reproducibility testing using Independent T-test results showed the difference between observers was not statistically significant (P = 0.275). Subjectively graded pterygium clinical grading based on its translucence appearance was repeatable and reproducible. Conclusion: These findings could serve as a basis for future work on to evaluate performance of pterygium clinical grading based on its morphology with different levels of experience and larger number of samples.
View
Kombinasi Jaringan Learning Vector Quantization Dan Normalized Cross Correlation Pada Pengenalan Wajah
Article
  • Feb 2020
Pengenalan wajah merupakan cara yang dilakukan untuk mengidentifikasi wajah berdasarkan nilai ciri yang terdapat pada citra wajah dan dapat diterapkan di dalam berbagai sistem, seperti absensi, akses keamanan ruangan dan login aplikasi atau perangkat. Salah satu algoritma untuk pengenalan wajah adalah LVQ (learning vector quantization), tetapi dalam pemilihan bobot awal yang kurang tepat dapat berdampak pada penurunan kinerja algoritma tersebut, sehingga hasil dari pengenalan wajah kurang akurat. Permasalahan ini dapat diselesaikan dengan penentuan bobot awal yang tepat dengan metode tertentu. Bobot yang dipilih pada penelitian ini berdasarkan kemiripan citra, salah satu metode untuk mengukur kemiripan adalah NCC (Normalized Cross Correlation). Penelitian ini akan dilakukan dengan mengkombinasi jaringan LVQ dengan menggunakan NCC dalam penentuan bobot awal untuk pengenalan wajah. Hasil pengujian yang diperoleh dengan kombinasi kedua metode tersebut untuk pengenalan wajah sebesar 94%.
View
Diagnostic value of nonlinear optical signals from collagen matrix in the detection of epithelial precancer
Article
Full-text available
  • Sep 2011
During precancer development in epithelium, neoplastic cells remodel the underlying stroma, for example, the basement membrane, capillaries, fibroblasts, and extracellular matrix. The purpose of this study is to investigate the relationship between the nonlinear optical signals from the collagen matrix in stroma and the progression of early epithelial carcinogenesis. Two-photon excited fluorescence (TPEF) and second harmonic generation (SHG) signals were measured from the stroma of hamster oral cheek pouch in vivo. We found that three features, including the intensity ratio of TPEF over SHG, the spatial frequency distribution, and the texture feature of SHG images, provide the quantitative identification of epithelial precancer at different pathologic stages. We demonstrated that the combination of all three features by using a support vector machine algorithm can significantly improve the accuracy in the detection of epithelial precancer.
View
An investigation of the textural characteristics associated with Gray Level Co-occurrence Matrix statistical parameters
Article
Full-text available
  • Apr 1995
The aim of this study was to investigate the statistical meaning of six GLCM (Gray Level Cooccurrence Matrix) parameters. This objective was mainly pursued by means of a self-consistent, theoretical assessment in order to remain independent from test image. The six statistical parameters are energy, contrast, variance, correlation, entropy and inverse difference moment, which are considered the most relevant among the 14 originally proposed by Haralick et al. The functional analysis supporting theoretical considerations was based on natural clustering in the feature space of segment texture values. The results show that among the six GLCM statistical parameters, five different sets can be identified, each set featuring a specific textural meaning. The first set contains energy and entropy, while the four remaining parameters can be regarded as belonging to four different sets. Two parameters, energy and contrast, are considered to be the most efficient for discriminating different textural patterns. A new GLCM statistical parameter, recursivity, is presented in order to replace energy which presents some degree of correlation with contrast. It is demonstrated that in some cases it may be reasonable to replace the computation of GLCM with that of GLDH (Gray Level Difference Histogram), in order to benefit by a better compromise between texture measurement accuracy, computer storage and computation time
View
Breast lesion classification on MRI by texture features
Article
  • Dec 2010
To explore the association between the texture feature and the lesions classification on breast MRI, we used a sixty-one biopsy-proven lesions database, using the Fisher stepwise discriminant analysis and logistic regression models.
View
GGO Nodule Volume-Preserving Nonrigid Lung Registration Using GLCM Texture Analysis
Article
  • Nov 2011
In lung cancer screening, benign and malignant nodules can be classified through nodule growth assessment by the registration and, then, subtraction between follow-up computed tomography scans. During the registration, the volume of nodule regions in the floating image should be preserved, whereas the volume of other regions in the floating image should be aligned to that in the reference image. However, ground glass opacity (GGO) nodules are very elusive to automatically segment due to their inhomogeneous interior. In other words, it is difficult to automatically define the volume-preserving regions of GGO nodules. In this paper, we propose an accurate and fast nonrigid registration method. It applies the volume-preserving constraint to candidate regions of GGO nodules, which are automatically detected by gray-level cooccurrence matrix (GLCM) texture analysis. Considering that GGO nodules can be characterized by their inner inhomogeneity and high intensity, we identify the candidate regions of GGO nodules based on the homogeneity values calculated by the GLCM and the intensity values. Furthermore, we accelerate our nonrigid registration by using Compute Unified Device Architecture (CUDA). In the nonrigid registration process, the computationally expensive procedures of the floating-image transformation and the cost-function calculation are accelerated by using CUDA. The experimental results demonstrated that our method almost perfectly preserves the volume of GGO nodules in the floating image as well as effectively aligns the lung between the reference and floating images. Regarding the computational performance, our CUDA-based method delivers about 20× faster registration than the conventional method. Our method can be successfully applied to a GGO nodule follow-up study and can be extended to the volume-preserving registration and subtraction of specific diseases in other organs (e.g., liver cancer).
View
An identification method of malignant and benign liver tumors from ultrasonography based on GLCM texture features and fuzzy SVM
Article
  • Oct 2010
  • EXPERT SYST APPL
In this study, we present an application of fuzzy support vector machine (FSVM) and image processing techniques for identifying liver tumor, including malignant and benign tumors. The gray level co-occurrence matrix (GLCM) matrices were utilized to evaluate the texture features of the regions of interest (ROI) of sonography in our experiment. Five textural features: energy, contrast, correlation, entropy, and homogeneity were extracted from the liver segmented images and analyzed using the texture average of four directions (0°, 45°, 90°, 135°) and distance, δ=6. The proposed system adopts the FSVM to distinguish between malignant and benign tumor cases more efficiently than support vector machine (SVM). The Gaussian RBF kernel has been used be more suitable for the application of identifying liver tumor from B-Mode ultrasound images than polynomial learning machine kernel and linear network kernel. The values of the parameters gamma (g) and regularization parameter (C) have been selected as 0.29 and 4.31×103, respectively. Via testing over 200 test cases by using RBF kernel, an overall accuracy of 97.0% has been received by the proposed FSVM algorithm. FSVM (AZ=0.984±0.014) obtain a better result than SVM (AZ=0.963±0.017) in recognition. It is demonstrated that the proposed FSVM algorithm and GLCM texture features technique are feasible and excellent in ultrasonography classification of liver tumor.
View
A Study on Using Texture Analysis Methods for Identifying Lobar Fissure Regions in Isotropic CT Images
Article
  • Sep 2009
The major hurdle for segmenting lung lobes in computed tomographic (CT) images is to identify fissure regions, which encase lobar fissures. Accurate identification of these regions is difficult due to the variable shape and appearance of the fissures, along with the low contrast and high noise associated with CT images. This paper studies the effectiveness of two texture analysis methods - the gray level co-occurrence matrix (GLCM) and the gray level run length matrix (GLRLM) - in identifying fissure regions from isotropic CT image stacks. To classify GLCM and GLRLM texture features, we applied a feed-forward back-propagation neural network and achieved the best classification accuracy utilizing 16 quantized levels for computing the GLCM and GLRLM texture features and 64 neurons in the input/hidden layers of the neural network. Tested on isotropic CT image stacks of 24 patients with the pathologic lungs, we obtained accuracies of 86% and 87% for identifying fissure regions using the GLCM and GLRLM methods, respectively. These accuracies compare favorably with surgeons/radiologists' accuracy of 80% for identifying fissure regions in clinical settings. This shows promising potential for segmenting lung lobes using the GLCM and GLRLM methods.
View
Breast lesion classification on MRI by texture features. 2nd ©2006-2015 Asian Research Publishing Network (ARPN) All rights reserved
  • Jan 2010
  • P Yachun
  • S Yuanzhi
  • L Li
Yachun P., Yuanzhi S. and Li L. 2010. Breast lesion classification on MRI by texture features. 2nd ©2006-2015 Asian Research Publishing Network (ARPN). All rights reserved. www.arpnjournals.com 6413