Early detection of glaucoma through retinal nerve fiber layer analysis using fractal dimension and texture feature

IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308
_______________________________________________________________________________________
Volume: 03 Issue: 10 | Oct-2014, Available @ http://www.ijret.org 158
EARLY DETECTION OF GLAUCOMA THROUGH RETINAL NERVE
FIBER LAYER ANALYSIS USING FRACTAL DIMENSION AND
TEXTURE FEATURE
Dharmanna Lamani1
, T. C. Manjunath2
, Mahesh M3
, Y. S. Nijagunarya4
1
Asst. Professor, Computer science & Engineering, SDMIT, Ujire, Karnataka, India,
2
Principal, Electronics & Communication, H.K.B.K, Bangalore, Karnataka, India,
3
Asst. Professor, Electronics & Communication, SDMIT, Ujire, Karnataka, India,
4
Professor, Computer Science & Engineering, Tumkur, Bangalore, India,
Abstract
The retinal nerve fiber layer (RNFL) is a vital part of human visual system, which can be directly observed by the fundus camera.
This paper describes a method for glaucomatous retina detection based on Texture and Fractal description, followed by
classification using support vector machine classifier. The color fundus images are used, in which the region of retinal nerve
fibers are analyzed. It is shown that Texture & Fractal dimensions are correlated and linear correlation coefficient values are
estimated at 0.35, 0.57, and 0.87 for healthy RNFL, medium loss and severe loss of RNFL respectively. The features are measured
at 303 RNFL regions retinal positions in the peri-papillary area from 50 non-glaucomatous and 24 glaucomatous retinal fundus
images. The presented method can also be used for glaucoma detection.
Keywords- Retinal nerve Fiber layer, Glaucoma, Fractal dimension, texture feature, Box counting method
-------------------------------------------------------------------***------------------------------------------------------------------
1. INTRODUCTION
Glaucoma is a second biggest retinal disease which leads to
permanent blindness worldwide. The world health
organization estimated 7.9 million people likely will get
affected by the year 2020. Loss of retinal nerve fiber layer is
the result of glaucoma disease. If not detected at initial
stages, the retinal nerve fiber layer loss becomes permanent
leading to blindness for the rest of life. Glaucoma cannot be
cured through surgical means or medicines. At the most,
ophthalmologist can provide treatment which helps to
slowdown the progression of RNFL loss. Hence, early
detection of RNFL loss is very essential for the
ophthalmologists, in order to initiate treatment. The RNFL
region is indicated as texture changes in color or grayscale
retinal fundus images. But there is no efficiently used
method for RNFL loss detection based on fundus images.
Although an escalating effort in this area is noticeable [2, 4,
and 5].
1.1 Background Work
The textural analysis of RNFL was performed on color
fundus images by R. Kolar et.al. and the analyzed results
were compared with RNFL thickness quantified by OCT.
Here, the authors have performed the analysis and compared
on normal subjects. This drives us to focus on defective
subjects [1].Textural analysis of RNFL was also performed
using Markov Random Fields. The atrophy of RNFL was
described through feature vector extracted from Causal
Autoregressive Random (CAR) model. A linear classifier
was used to classify the defected RNFL and normal object
with a classification error of below 4%.[2]. Fractal
descriptions can also be a measure to detect RNFL loss. [4].
A model was built by M.A. Mayer et.al. using RNFL
thickness and textural features extracted through
segmentation [7]. Another kind of textural analysis was
performed by A.Novatny et.al. which was based on local
binary patterns and Gaussian Markov Random fields. In this
work, RNFL was quantified and a classification error of
about 3% was achieved [6]. A Glaucoma Risk Index (GRI)
was introduced by R. Bock et.al. by combining a
probabilistic 2-stage classifier and an appearance-based
dimension reduction technique [17]. The quantification of
RNFL was also done using three SD-OCT (Spectralis,
Cirrus and RTVue) and one TD-OCT (Stratus). Despite high
correlation among these instruments, RNFL values were
significantly different between the instruments [16].
The identification and quantification of RNFL was
simplified using Confocal Laser scanning Microscopy
(CLSM) which provided enface and cross-sectional images
of a whole mounted retina. [18]. E.Z Blumenthal et.al.
assessed RNFL qualitatively using ophthalmoscopy, color
stereophotography and re-free monochromatic photography.
The quantification of RNFL was also done using Confocal
Scanning Laser Ophthalmoscopy (CSLO), nerve fiber layer
analyzer, OCT and Retinal Thickness Anayzer (RTA) [14].
Dhivyabharathi et.al. used Gabor filter for texture
segmentation through which various layers were extracted
from OCT images [9]. The algorithm proposed by the author
segmented the region of interest efficiently without human
intervention. This led to the measurement of RNFL
thickness. These results were compared with those obtained

_______________________________________________________________________________________
from the hospitals. The appearance of speckle noise is due to
the coherent processing of backscattered signals from
multiple distributed targets. Speckle noise reduces quality
and diagnostic usefulness of OCT images. Due to these
issues, the interpretation of images will face difficulties.
Also, edge plays an important role in identifying the layers
present in an image. Hence, the authors finalized to use a
median filter since it is characterized by edge preserving
nature.
Cup-to-Disc ratio (CDR) is also an important parameter
used in the diagnosis of glaucoma. CDR was always
determined manually by trained clinicians. Hence C.B.
Anusorn et.al. calculated CDR automatically from fundus
images. To extract cup, color component analysis and
threshold level-set method were evaluated. Whereas edge
detection method and variational level-set method were
used. An accuracy of 89% was recorded when these results
were compared with those of clinical data [11].
The evaluation of retina through fundus camera is subjective
and so efficiently reproducible. Whereas OCT allows
quantification of RNFL directly and moreover the diagnostic
procedures through OCT are reproducible. Apart from these
merits, OCT diagnoses are expensive due to which they are
not so widely available at many ophthalmic clinics. Hence,
J. Odstrcilik et.al. have considered the faster and cheaper
image acquisition using fundus camera. Through subjective
analysis it was believed that variations in RNFL thickness
led to changes in visual appearance of RNFL pattern. Later
it was learnt that even in healthy retina, the variations of
RNFL thickness anatomically was present which depended
on the angular position around the ONH. These findings
encouraged the authors to carry out analysis of RNFL
variations through advanced image processing methods [2].
Hence we conclude that fundus camera images provide
useful and easy-to-access information about RNFL which
helps us to assess glaucoma through GRI or C/D ratio.
This paper describes a novel method to classify healthy
RNFL, medium loss and severe loss RNFL using fractal
dimension & texture feature properties. Section 2 presents
the material and methodology with respect to RNF analysis.
Section 3 briefly describes the pseudo-code and texture
properties, & also discussed is the selection of features.
Section 4 describes the result and discussion based on
texture feature and Fractal Dimensions. Finally conclusion
is put up in section 5.
Fig 1 Retinal image from database showing macula, optic
disk, blood vessels and region with/without retinal nerve
fibers.
One of the images from database is shown in fig.1. It shows
various regions of fundus. Region 1 is called macula which
is oval in shape and is made up of two or more layers of
ganglion cells. Regions 2 and 3 show healthy RNF layers.
These regions are brighter as compared to regions 4 and 5.
The latter two regions which are dark in nature show RNFL
loss. Region 7 is the optic disk where blood vessels are
accumulated. Also RNFL radiate from optic disk towards
the macula. It is the RNFL which forms the optic nerve
while leaving optic disk towards brain. The dark appearance
of defected region of RNFL is due to its atrophy. The simple
physics behind this is lowered reflection of incident light.
The dark nature of RNFL in certain areas is not a reliable
factor to conclude on glaucoma. Fig. 2 shows the fundus
image in which certain areas enclosed by black squares
depict healthy RNFL region. The white squared regions
show RNFL losses. The textural features were extracted
from these areas for analysis. A total of 101 healthy regions,
101 medium loss and 101 severe loss in RNFL were
extracted from a set of 50 non-glaucomatous and 24
glaucomatous retinal fundus images.
Fig 2 Detail of the fundus image. The dark squares show the
samples taken for the fractal and texture features analysis
from area of healthy RNF. The white squares show the
samples taken for the fractal & texture features analysis
from area of loss of the RNF.

_______________________________________________________________________________________
2. METHODOLOGY
2.1 Material Used
The fundus images were collected from Shri Dharmasthala
Manjunatheshwara Hospital, Dharwad (SDMHD),
Karnataka, India. All images were captured with a
resolution of 530 X 720 pixels. For experimentation, we
have used 200 retinal images: 100 non-glaucoma and 100
RNFL loss images obtained from individuals of age group
25-56 years. Doctors in Ophthalmology department of the
hospital certified image quality and usability. Fig. 1 & Fig. 2
represent the typical healthy (non-glaucoma) and RNFL
damaged region retinal images. The fundus camera with
microscope and light source were adopted to capture retinal
images in order to diagnose glaucoma, neo-vascular
glaucoma, macular degeneration, diabetes, etc.
2.2 Image Pre-Processing
Image pre-processing can significantly increase the
reliability of a small neighborhood of a pixel in an input
(I/p) image to get a new brightness value in the output (O/P)
image. Such image pre-processing step consists of an image
using normalization operator. Image normalization
significantly increases the dynamic range of the histogram
of the image. The intensity value in an input image includes
a uniform distribution of intensity. As a result, the contrast
of the image was increased. In second pre-processing step,
Retinal nerve fiber layer loss retinal images are darker as
compared to healthy RNFL retina. Hence we highlighted red
channel component in the image for computation of fractal
dimension.
2.3 Extracting Image Features
The extraction of image feature is explained through the
flow chart shown in Figure 3.
Fig 3 Flow Chart for Extraction of Image Feature
Fig 4 Box Counting Principle
2.5 Fractal Dimension
Fractal dimension allows us to measure the complexity of an
object, that too preferably self similarity such as eye, brain,
tree, mountain and so on so forth. Recently, Hsiao- Wen
Chung et.al. presented fractal analysis approach for Nuclear
medicine images for the diagnosis of Pulmonary
Emphysema and concluded that FD seems to be an over
simplified parameter unrelated to spatial heterogeneity of
images [18]. Fractal analysis is potentially suitable for an
objective quantification of spatial heterogeneity because it is
believed to be effective in characterizing complex systems
that are hard to describe using conventional Euclidean
geometry. Fractal dimension is a parameter that can be
determined solely by the image intensity [16].
2.5 Box Counting Method (BCM)
Box counting method is one of the most popular algorithms
to determine fractal dimension of an image. Box counting
method is described with the following example. The image
was resized to gray scale image and was converted into
binary image by bw method later to create square boxes,
binary image was resized to a nxn so that the length,
measured in number of pixels of a power of 2 this helps for
the square image to be equally divided in to four sub parts
and again each subparts divided in to four subparts and so
on. The number of pixel which contains “white” pixels was
recorded as function of the box size N(s), length of box. The
reciprocal of box size is also noted as 1/r, this principle is
repeated for all the binary images obtained by different
threshold values ranging from 0.2 to 0.5 with interval 0.1 of
an average binary retinal image. From figure 4, Number of
boxes N(s) =4 number of boxes which contains ones N (1/s)
=3. Hence fractal dimension =1.333.
2.6 Algorithm for Box Counting Method
The box-counting method was used on retinal images to
exhibit a fractal structure. The image was resized to measure
(256 × 256) and then converted into gray scale. Grayscale
images became binary images with a threshold of 0.2 to 0.5
with distance of 0.1, meaning that if the pixel had a binary
value greater than threshold (on a 0 to 1 scale) the pixel
became “black”, or assigned the value of 0, otherwise if the
pixel becomes „white‟, then a value of 1 is assigned. The
Algorithm developed for the box counting method as given
below.1.26.(G) Texture analysis of RNFL in retinal images
Various fractal dimension techniques were applied in our
previous work [21, 22, 23, and 24]. Texture methods &
statistical based techniques are common mathematical tools

_______________________________________________________________________________________
for texture classification. Hence these tools are employed for
RNFL analysis. There are mainly three types of statistical
techniques: first order statistics, second order statistics, and
higher order statistics. In this paper, first order statistics is
adopted, which depends on individual pixel intensity values,
but not on pixels with inter relationship. The interpretations
of parameters obtained through first order statistics are
straight forward & provide a view on textural properties and
its visual appearance. The various parameters of first order
statistics are standard deviation, average, skewness,
kurtoses, and shannon entropy etc. Among these parameters,
mean and standard deviation are computed using intensity
probability distribution (IPD), which is calculated from
histogram of ROI. The detailed explanations of these
parameters can be found elsewhere [1, 2]. The equations for
parameters used in this paper are given in Table-1.
Table – 1: Mathemtical equation for Mean and SD
SL
No
Features Equation Description
1 Mean 

N
i
iiHu
0
)(
H(i) denotes a
PDF and is
computed from
histogram.
H(i)=ni/N,
where ni is the
ith pixel for
i=0,1,2….L-1,
and N is the
total number
pixels. L is 255
(Gray levels).
2 SD
)()(
21
0
iHi
N
i



 
3. RESULT & DISCUSSION
The texture and fractal dimension features were extracted
from each image sample: two features were estimated for all
samples from each group (X, Y, and Z). The size of sample
is 32 X 32 pixels. Which has been selected a sufficiently
large region with RNFL striation. The upper limit of sample
size was forbidden by the blood vessels & several other
anatomical structures in the retinal images. First, the
correlation coefficient is estimated for investigation. It is
shown that Texture & Fractal dimensions are correlated and
linear correlation coefficient values are estimated at 0.87,
0.57, and 0.35 for healthy RNFL, medium loss and severe
loss of RNFL respectively, result shown in table 3. The
result of correlation coefficient is put up in table 2, showing
their linear dependency. The configuration was established
based on hypothesis test with p-value < 0.01. In table-2,
sample A and B belong to class X (severe loss of RNFL
dataset). The sample C and D belongs to class Y (medium
loss RNFL dataset). Finally, a sample E and F belongs to
class Z (healthy RNFL dataset). Also, in table-4, statistical
data such as mean, median, standard-deviation etc for fractal
dimension and texture feature are depicted.
Figure 5 shows the plot of fractal dimension versus
logarithmic values of texture features. Fractal dimension
ranges from 1.01 to 1.72 and logarithmic mean texture
feature ranges from 0.3 to 1.7 for dataset for class X. Fractal
dimension ranges from 1.73 to 1.8399 and logarithmic mean
texture feature ranges from 1.72 to 1.9 for dataset for class
Y. Fractal dimension ranges from 1.84 to 1.99 and
logarithmic mean texture feature ranges from 1.9 to 2.3 for
dataset for class Z. Figure 6 describes sequence of operation
performed to detect RNFL loss from region of interest of
retinal image. Figure 6(a) shows RNFL loss retinal image
and figure 6(b) is sample of the RNFL loss which is
considered for the estimation of mean texture feature and
FD. Mean texture feature is computed by equation 1 which
is given in table 1. Whereas fractal dimension estimated
from figure 6(c) that is 1.6357.
Fig. 5: Graph of FD versus logarithmic mean of ROI
(ln(µ))
(a)
(b)
FD=1.6357, µ =47.7477) , ln(µ =2.34
(c)
Fig. 6: (a) RNFL loss retinal images, (b) RNFL loss ROI
and mean µ=47.7477, and (c) Fractal Dimension 1.6357

_______________________________________________________________________________________
Table 2: The table summarizes the Spearman‟s correlation coefficients computed from samples in particular image. The mean
value and fractal dimension values are presented. (Sample A & B belongs to Class X, Sample A & B belongs to Class Y, and
Sample A & B belongs to Class Z).
A B C D E F
FD 1.4256 1.6734 1.7234 1.8091 1.8450 1.8444
µ
52.2525 66.4525 81.4521 116.5927 123.3822 118.6125
ln(µ)
1.7281 1.8225 1.9092 2.0666 2.0912 2.0721
Table 3: Spearman correlation coefficients between texture and fractal dimension features for the whole dataset; p value < 0.01
Classes Class X Class Y Class Z
Correlation 0.35 0.57 0.87
Table 4: Statistical data shown for ROI of healthy RNFL tissue, Medium RNFL loss, and severe RNFL loss.
SL. No Attributes
Healthy RNFL
(class X)
Medium loss
(class Y)
Severe loss
(class Z)
FD µ FD µ FD µ
1 Mean 1.465 1.805 1.817 1.921 1.894 2.13
2 SD 0.2473 0.1021 0.03075 1.869 0.04515 0.04954
3 Range 0.8276 0.4867 0.1574 0.3299 0.1661 0.2535
4 Median 1.51 1.794 1.825 1.869 1.88 2.131
5 Mode 1.012 1.608 1.845 1.812 1.844 2.011
6 Min 1.02 1.608 1.716 1.812 1.821 2.011
7 Max 1.839 2.094 1.873 2.142 1.987 2.264
4. CONCLUSION
In this paper, authors presented a solution for retinal nerve
fiber layer loss estimation using fractal dimension & texture
feature, in order to early detection of glaucoma progression.
This method is contrast to earlier approaches which have
mainly considered on the computations of cup-disk ration
(CDR) which varies considerably within normal‟s and also
detects glaucoma after 40% RNFL loss. The presented work
enables more comprehensive assessment of the retinal
nerve fiber layer and performing glaucoma detection using
RNFL loss determining through fractal dimension and
texture feature.
Spearman correlation co-efficient estimated for FD and
texture feature for healthy, medium loss, and severe loss
RNFL are 0.85, 56, and 35 respectively. The proposed
features can be used as a part of feature vector in glaucoma
risk, as explained in [1]. These features may be also used in
the glaucoma screening program together with other features
such as CDR, ISNT ration, and RNFL thickness etc. based
on different methods [5, 6, and 7].
REFERENCES
[1]. Radim Kolar, Ralf P. Tornow, Robert Laemmer, Jan
Odstrcilik, Markus A. Mayer, Jiri Gazarek, Jiri Jan,Tomas
Kubena, and Pavel Cernosek, “Analysis of visual
appearance of retinal nerve fibers in high resolution fundus
images: a study on normal subjects,” Hindawi Publishing
Corporation Computational and Mathematical Methods in
Medicine Volume 2013, Article ID 134543,pp 1-10.
[2]. Jan Odstrcilika,b,∗, Radim Kolara,b, Ralf-Peter
Tornowc, Jiri Janb, Attila Budaid,Markus Mayerd, Martina
Vodakovab, Robert Laemmerc, Martin Lamosb, Zdenek
Kunab,Jiri Gazarekb, Tomas Kubenae, Pavel Cernoseke,
Marina Ronzhina, “Thickness related textural properties of
retinal nerve fiber layer in color fundus images,”
Computerized Medical Imaging and Graphics,Vol. 38, 2014,
pp. 508–516.
[3]. Radim Kolar, Jiri Jan, “Detection of Glaucomatous Eye
via Color Fundus Images Using Fractal Dimensions,” Radio
engineering, vol. 17, no. 3, pp. 109-114, September 2008.
[4]. R. Kolar and P. Vacha, “Texture Analysis of the Retinal
Nerve Fiber Layer in Fundus Images via Markov Random
Fields,” IFMBE Proceedings 25/XI, pp. 247–250, 2009.
[5]. Prof. A. Annadhason, “Methods of fractal dimension
computation,” IRACST - International Journal of Computer

_______________________________________________________________________________________
Science and Information Technology & Security (IJCSITS),
ISSN: 2249-9555,Vol. 2, No. 1, pp. 167-169, 2012.
[6]. Novotný A., Odstrčilík J., Kolář R., Jan J., “ Texture
Analysis Of Nerve Fibre Layer In Retinal Images Via Local
Binary Patterns And Gaussian Markov Random Fields,”
BIOSIGNAL, Analysis of Biomedical Signals and Images;
Vol. 20: pp. 308-315, 2010.
[7]. Markus A. Mayer, Joachim Hornegger, Christian Y.
Mardin, and Ralf P. Tornow, “Retinal Nerve Fiber layer
segmentation on fd-oct scans of normal subjects and
glaucoma patients,” Biomedical optics express, Vol. 1, No.
5. pp. 1-26, Nov. 2010.
[8]. Carol Yim Lui Cheung, Christopher Kai-shun Leung,
“A Practical Guide for Interpretation of Optical Coherence
Tomography Retinal Nerve Fiber Layer Measurement,”
Journal of Current Glaucoma Practice, January-April 2009;
Vol. 1, No. 3, pp.9-13.
[9]. Dhivyabharathi, Ganeshbabu2 and Venkatesh Rengaraj,
“Measurement of rnfl thickness using oct images for
glaucoma detection,” ICTACT journal on image and video
processing, ISSN: 0976-9102, August 2013, volume: 04,
issue: 01, pp. 637-641.
[10]. Dharmanna Lamani, Dr. T. C. Manjunath, Dr. U P
Kulkarni, “Automated detection of neovascular glaucoma
through fractal dimension method,” International Journal of
Computer Science and Information Technologies, ISSN:
0975-9646, Vol. 5 (4) , 2014, pp. 5252-5257.
[11]. Chalinee Burana-Anusorn, Waree Kongprawechnon,
Toshiaki Kondo, Sunisa Sintuwong and
KanokvateTungpimolrut, “Image Processing Techniques for
Glaucoma Detection Using the Cup-to-Disc Ratio,”
Thammasat International Journal of Science and
Technology, Vol. 18, No. 1, January-March 2013, pp 22-34.
[12]. Ingrida Januleviˇciene, Rita Ehrlich, Brent Siesky,
Irena Nedzelskien˙e, and Alon Harris, “Evaluation of
Hemodynamic Parameters as Predictors of Glaucoma
Progression,” Hindawi Publishing Corporation Journal of
Ophthalmology, Volume 2011, pp 1-9.
[13]. Hsuan Tiem Lin, “Texture classification using Fractal
based features and support vector machines,” National
Taiwan University, Taipei 106, Taiwan. PP. 1-22.
[14]. Eytan Z. Blumenthal, MD, and Robert N. Weinreb,
MD, “Assessment of the Retinal Nerve Fiber Layer in
Clinical Trials of Glaucoma Neuroprotection,” Survey of
ophthalmology volume 45, No 3, May 2001, pp 305-312.
[15]. Roge´rio A. Costa, Mirian Skaf, Luiz A.S. Melo Jr.,
Daniela Calucci, Jose A. Cardillo, Jarbas C. Castro, David
Huang, Maciej Wojtkowski, “Retinal assessment using
optical coherence tomography,” Progress in Retinal and Eye
Research, Vol. 25, 2006, pp. 325–353.
[16]. Leonard K. Seibold, Naresh Mandava, And Malik Y.
Kahook, “Comparison of Retinal Nerve Fiber Layer
Thickness in Normal Eyes Using Time-Domain and
Spectral-Domain
Optical Coherence Tomography,” American journal of
ophthalmology, Vol. 150, NO. 6, pp. 807-814, Dec. 2010.
[17]. Rüdiger Bock a,b,*, Jörg Meier a, László G. Nyúl c,
Joachim Hornegger a,b, Georg Michelson, “Glaucoma risk
index:Automated glaucoma detection from color fundus
images,” Medical Image Analysis, Vol .14, 2010, pp. 471–
481.
[18]. Xiang-Run Huang*, Robert W. Knighton, Valery
Shestopalov, “Quantifying retinal nerve fiber layer thickness
in whole-mounted retina,” Elsevier Experimental Eye
Research Vol. 83, 2006, pp. 1096-1101.
[19]. Thomas Kersey*, Colin I. Clement**, Phillip Bloom*
& M. Francesca Cordeiro,“ New trends in glaucoma risk,
diagnosis & management,” Indian J Med Res 137, April
2013, pp 659-668.
[20]. Amineh. Naseri, Ali.Pouyan, Nader.Kavian,
“Automatic Detection of Retina Layers using Texture
Analysis,” International Journal of Computer Applications,
ISSN: 0975 – 8887, Volume 46– No.1, May 2012. pp 29-33.
[21]. Dharmanna Lamani, Dr. T.C.Manjunath, “Automatic
Detection of Diabetes Retinopathy through Image Feature
Power Spectral Fractal Dimension ”, 8th Elsevier
International Conference on Image and Signal Processing
(ICISP-2014),UVCE, July 25th-26th , Bangalore,
Karnataka, India, Paper ID : ICISP-357, Reference
No.: 357I-S,
[22]. Dharmanna Lamani, Dr. T.C.Manjunath, Ramegowda,
“Fractal dimension with object rotation : A case study with
glaucoma eye”, 5th IEEE Int. Conf. on Signal & Image
Processing (ICSIP-2014), BNMIT, Bangalore, Karnataka,
India, 978-0-7695-5100-5/13 $31.00 © 2013 IEEE DOI
10.1109/ICSIP.2014.22, ISBN 978-0-7695-5100-5, Paper
No. ICSIP-31, Session-Medical Image Processing, pp. 111-
116, 8-10 Jan. 2014.
[23]. Dharmanna Lamani, Dr. T.C.Manjunath,“Automated
Detection Of Diabetic Retinopathy Through Image Feature
Extraction”, 2014 IEEE International Conference On
Advances In Electronics, Computers And Communications
(Icaecc), OCTBER 10-11-2014, Bangalore, Karnataka,
India.
[24]. Dharmanna Lamani, Dr. T.C.Manjunath, Ramegowda,
“Cup-disk segmentation and fractal dimension to detect
glaucomatous eyes”, National Conf. on Current Trends in
Research & Technology (CRT-13), Sponsored by IET (The
institute of engineering and technology, UK), SDMIT, Ujire,
South Kanara, Karnataka, India, pp. - , 27 to 28 Sep. 2013.

Early detection of glaucoma through retinal nerve fiber layer analysis using fractal dimension and texture feature

More Related Content

What's hot (20)

Viewers also liked (17)

Similar to Early detection of glaucoma through retinal nerve fiber layer analysis using fractal dimension and texture feature (20)

More from eSAT Journals (20)

Recently uploaded (20)

Early detection of glaucoma through retinal nerve fiber layer analysis using fractal dimension and texture feature