Study on Glaucoma Detection Using CNN

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 06 | June 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 3302
Study on Glaucoma Detection Using CNN
Jayraj F1, Aditya K2, Mahit M3, Nihal S4, Rahul K5, Pushpalatha S.Nikkam6
1Department of Information Science and Engineering, SDMCET, Dharwad, Karnataka, India
2,3,4,5Student, Department of Information Science and Engineering, SDMCET, Dharwad, Karnataka, India
6 Pushpalatha S. Nikkam , 6Assistant Professor, Department of Information Science and Engineering, SDMCET,
Dharwad, Karnataka, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Glaucoma is a persistent and incurable eye
condition that makes eyesight and life quality to deteriorate.
We present a deep learning (DL) architecture using a
convolutional neural network for automated glaucoma
detection in this study. Deep learning systems, like as CNN
models, can infer a hierarchical representation of pictures in
order to distinguish between glaucoma and non-glaucoma
patterns for diagnostic purposes. Six learned layers are
included in the proposed DL architecture: four convolutional
layers and two fully-connected layers. In this paper, we
suggest a CNN method to glaucoma diagnosis. We create a
network using Convolutional Neural Network (CNN)
architecture and data augmentation to recognize the subtle
elements involved in the classification job, such as
microaneurysms, exudate, and haemorrhages on the retina.
Key Words: Convolutional Neural Network, Deep
Learning.
1. INTRODUCTION
Glaucoma is a disorder that damages the optic nerve in
your eye and worsens over time. It is frequently linked to
greater in eye pressure. Glaucoma is usually hereditary
and may not manifest it until later in life. The increased
pressure, known as optic nerve, which sends images to the
brain, can be damaged by intraocular pressure. Glaucoma
can cause irreversible vision loss if the damage also isn't
treated. Glaucoma, if left untreated, can result in complete
and irreversible blindness within several years.
1.1 Existing System
Glaucoma is frequently detected too late because it is
generally asymptomatic for years: half of all cases
have mitigate to progressive disease first shows itself,
it is in the worse eye, even in countries with high
standards, massively increasing the disease's human
and economic burden on individuals and society.
Measurements of intraocular pressure (IOP) are used
during regular eye exams, but they cannot distinguish
between healthy and glaucomatous eyes up to half of
glaucoma patients may not have an elevated IOP upon
examination, and many ocular hypertensive patients
do not require treatment and will never develop
glaucoma. By several observational studies
(Rotterdam Eye Study, Blue Mountains Eye Study,
Visual Impairment Project, Proyecto VER, and Latino
Eye Study), glaucoma is untreated in 50 percent of
cases in the Western part of the world, with greater
rates in specific ethnic groups, and up to 90 percent in
developing countries. On the other side, glaucoma is
frequently over treated: many patients are treated
even when they have no illness. This strongly
advocates for a global increase in illness detection
precision.
1.2 Proposed System
Glaucoma is an eye disorder that leads to lifelong
blindness. Glaucoma is a chronic condition that can
only be prevented if it is recognised properly at an
early stage. The proposed method creates an
automated glaucoma detection computer-aided
system that allows ophthalmologists to accurately
diagnose glaucoma patients early. The method uses a
pre-processed fundus picture and extracts the optic
cup and optic disc before calculating the Cup to disc
ratio. To train and evaluate the classifier, intensity and
textural information are taken from the picture. The
outcomes of disease diagnosis using CDR are
combined with characteristics to identify the picture
as glaucoma or non-glaucoma suspect.
2. DESIGN AND DEVELOPMENT
2.1 Objectives
 To develop an efficient Pre-processing /Data
Augmentation technique.
 To develop a Novel algorithm to extract features
using CNN.
 To develop high computational classifier to detect
the Glaucomatous images

2.2 Tools and Technologies
 Pycharm IDE development of code.
 Python version 3.7 for implementation of
algorithms.
 Convolutional neural network for Glaucoma
Detection.
 TensorFlow- TensorFlow is a complete open
source machine learning framework.
 FrontEnd: HTML5, CSS3, JavaScript.
2.3 Methodology
By processing features extraction and classification
simultaneously inside the same network of neurons, we
construct an algorithm using CNN architecture that avoids
the standard hand-crafted features extraction stage and, as
a result, automatically and without user input provides a
diagnosis. One of the most sophisticated picture
categorization models available today is the Convolutional
Neural Network. They are divided into two parts. An image
in the form of a pixel matrix is provided as input. It is a gray
scale image with two dimensions. To represent the
fundamental colors, a third dimension, depth 3, is used
(RGB). The first section of a CNN is the traditional
component. It works as an image characteristic extractor.
Convolution maps are created by passing a picture through
a sequence of filters, or convoluted nuclei. Some
intermediate filters use a local maximum operation to
lower image resolution. Eventually, the convolution maps
are flattened and concatenated to generate a CNN code,
which is a feature vector. CNN obtains this code from the
evasive party, which is then linked in the entrance of a
second component, which is made up of totally connected
layers (multilayer perceptron). The purpose of this section
is to combine the CNN code's features in order to
categorize the image. As a result, the final layer has one
neuron of each type. Pre-processing is used in the input
layer to shrink photos with random pixels to 224 × 224 x 3.
The picture pixels 224 x 224 x 3 are then convolved using
the weights and bias terms in the convolutional layer. The
activation function (Rectified linear) is used to convert all
negative values produced after convolution to 0 while
maintaining positive values unchanged. This activation
function determines to choose whether or not trigger a
neuron. The data are then passed to the max-pooling layer,
which decreases the image dimensionality and boosting
computing performance. Again, this pooled image is
provided for batch normalization, which rationalizes the
image in each channel within 0 and 1. (RGB). And this
procedure is repeated three times with the layer network
in the important difference, the dropout periodically
dropping certain units in our model to reduce model
complexity and optimize performance speed. We also use a
regularizes that averages all the squares weighting factors
in the weight matrices to generate the gradient descent,
which updates the weights to reduce model loss. Following
all of these steps, the picture pixel values are sent to the
flattening layer, which aids in the transformation of 2D
data to ID for input to the fully - connected network for
classification.
A data-flow diagram (DFD), like the one in Figure 1,
demonstrates how data flows through with a process or
system. The DFD also give more knowledge about each
entity's inputs and outputs as well as the process itself.
Fig-1: Data Flow Diagram of the system.
3. RESULTS AND DISCUSSIONS
Most current supervisory algorithms allow for additional
pre- or post-processing stages in order to differentiate
between the various stages of glaucoma. Further
algorithms that demand human feature extraction stages
are required to specify the fundus images. Convolutional
Neural Networks (CNN) provide a comprehensive solution
to all stages of glaucoma in our suggested solution. There
is no need for manual feature extraction procedures. With
dropout techniques, our network architecture produced a
significant improvement in classification accuracy. Our
network architecture is complex and computationally
demanding, requiring the need for a graphics processing
unit to process fundus images as the number of layers is
increased. By increasing the number of photos in each
class and the number of convolutional layers, it is possible
to improve the testing accuracy for Glaucoma.

A thorough explanation of the Convolutional Neural
Network (CNN) was provided in this project. It was
explained how different layers, including convolution,
polling, Rectified Linear Unit (ReLu), and fully connected
layers, function. We are known that the pooling layer
reduces dimensionality, the ReLU layer tends to increase
non-linear properties, and the fully connected layer is the
result of the previous layer. The convolution layer is used
to extract features from the input image. To put it another
way, the fully connected layer takes neurons from the
previous layer and connects them to every single neuron it
has to form a neural network, which will then be gathered
for further classification.
Input:
Image of human eye is put into the glaucoma detector
system as shown in figure 2 to check if it is healthy or not.
Fig2:Browsing human eye Images
Output:
Fig-3: Healthy Eye
If it is healthy as shown in figure 3
Fig-4:Glaucoma affected Eye
If it is affected by glaucoma as shown in figure 4
Fig 5: Model Accuracy and Model loss
The graph as shown in figure 5 describe the model
accuracy and model loss in which we compare the
validated data with trained data and the accuracy.
4. CONCLUSIONS
Convolutional Neural Network (CNN) is a systematic
approach to all levels of Glaucoma in our proposed
solution. There are no manual feature extraction processes
required. The classification performance of our network
design using dropout methods was substantial. Our
system architecture is complicated and computationally
costly, necessitating the use of a graphics processing unit
to interpret fundus pictures as the number of layers piled
increases. The accuracy of glaucoma testing can be
improved by increasing the number of images in each
class and the number of fully connected layers.
ACKNOWLEDGEMENT
We have been given the privilege of thanking everyone
who assisted us in the completion of the paper. We'd want
to offer our heartfelt appreciation to Dr. Pushpalatha S.
Nikkam and Dr. Jagadeesh Pujari of the Department of
Information Science and Engineering at SDMCET
Dharwad, our project guide and Head of the Department
for assisting and guiding us throughout the process of

polishing the work responsible for producing this paper.
Finally, we owe a great deal to our parents for their
unwavering support and assistance.
REFERENCES
[1] R. R. Bourne, H. R. Taylor, et al.,“Number of people
blind or visually impaired by glaucoma worldwide and in
world regions 1990–2010: a meta-analysis,” PloS one, vol.
11,no. 10, p. e0162229, 2019.
[2] G. Litjens, T. Kooi, et al.“A survey on deep learning in
medical image analysis,” Medical image analysis, vol. 42,
pp. 60–88, 2020
[3]. Yousefi, M. H. Goldbaum, et al. “Glaucoma progression
detection using structural retinal nerve fiber layer
measurements and functional visual field points,” IEEE
Trans. Biomed. Eng., vol. 61, no. 4, pp. 1143–1154, Apr.
2021.
[4] S. S. Kanse and D. M. Yadav. "Retinal fundus image for
glaucoma detection". A review and study. Journal of
Intelligent Systems, 28(1):43–56, 2021
[5] Q. Abbas, A. Mateen, et al. “Glaucoma-deep: detection of
glaucoma eye disease on retinal fundus images using deep
learning,” Int J Adv Computer Sci Appl, vol. 8, no. 6, pp. 41–
5, 2021.
[6] X. Chen, Y. Xu, et al. “Glaucoma detection based on deep
convolutional neural network,” in 2015 37th annual
international conference of the IEEE engineering in
medicine and biology society (EMBC). IEEE, 2020
[7] A. Diaz-Pinto, S. Morales, et al. “Cnns for automatic
glaucoma assessment using fundus images: an extensive
validation,” Biomedical engineering online, vol. 18, no. 1, p.
29, 2019.
[8] S.M. Nikam and C.Y. Patil. "Glaucoma detection from
fundus images using MATLAB gui". In 2017 3rd
International Conference on Advancesin Computing,
Communication & Automation (ICACCA)(Fall), pages 1–4,
2021.
[9] Michael H. Goldbaum, Madhusudhanan
Balasubramanian, et al. “Learning from Data: Recognizing
Glaucomatous Defect Patterns and Detecting Progression
from Visual Field Measurements, vol. 8, no. 6, pp. 41–5,
2021.
[10] Juan Carrillo, Lola Bautista, et al. "GLAUCOMA
DETECTION USING FUNDUS IMAGES OF THE EYE,” in
2015 37th annual international conference of the IEEE
engineering in medicine and biology society (EMBC). IEEE,
2020
BIOGRAPHIES
Jayraj Fasalkar, Student,
Department of Information Science
and Engineering, SDMCET,
Dharwad, Karnataka, India
Aditya Singh, Student, Department
of Information Science and
Engineering, SDMCET, Dharwad,
Karnataka, India
Mahit Mokashi, Student,
Department of Information Science
and Engineering, SDMCET,
Dharwad, Karnataka, India.
Nihal Shetty, Student, Department
Karnataka, India
Rahul Kumar, Student, Department
Karnataka, India
Pushpalatha S. Nikkam, Assistant
Professor, Department of
Information Science and
Karnataka, India

Study on Glaucoma Detection Using CNN

Recommended

More Related Content

Similar to Study on Glaucoma Detection Using CNN (20)

More from IRJET Journal (20)

Recently uploaded (20)

Study on Glaucoma Detection Using CNN