VISVESVARAYA TECHNOLOGICAL UNIVERSITY

Belagavi, Karnataka, India

Mini Project Report on

“RECOGNITION OF DIABETIC RETINOPATHY
USING CNN ALGORITHM”
Submitted in partial fulfillment of the requirement of VI Semester
Mini Project (18ECMP68)
Submitted by,
AKSHAT GUPTA 1RF20EC003
MOHAMMED 1RF20EC028
NADEEM SHIVAM 1RF20EC042
KUMAR RAI 1RF20EC053
VIKRANT RANA
Under the guidance of Dr. KAVITHA N,
Assistant Professor,
Dept. of ECE, RVITM

DEPARTMENT OF
ELECTRONICS & COMMUNICATION
ENGINEERING
®

RV INSTITUTE OF TECHNOLOGY AND MANAGEMENT®

BANGALORE-560076
2022-23
 RV INSTITUTE OF TECHNOLOGY AND MANAGEMENT®
(Affiliated to Visvesvaraya Technological University, Belagavi & Approved by AICTE, NewDelhi)
Bengaluru-560076

DEPARTMENT OF
ELECTRONICS AND COMMUNICATION ENGINEERING

CERTIFICATE
Certified that the project work titled “RECOGNITION OF DIABETIC
RETINOPATHY USING CNN ALGORITHM” is carried out by AKSHAT
GUPTA(1RF20EC003),MOHAMMED NADEEM(1RF20EC028), SHIVAM
KUMAR RAI(1RF20EC042), and VIKRANT RANA(1RF20EC053), who are
bonafide students of RV Institute of Technology and Management, Bangalore, in partial
fulfillment for the award of degree of Bachelor of Engineering in Electronics and
Communication Engineering of the Visvesvaraya Technological University, Belagavi during
the year 2022-2023. It is certified that all corrections/suggestions indicated for the internal
Assessment have been incorporated in the report deposited in the departmental library. The
project report has been approved as it satisfies the academic requirements in respect of project
work prescribed by the institution for the said degree.

Signature of Guide Signature of Head of the Department Signature of Principal

Dr. Kavitha N Dr. Prashant P Patavardhan Dr. Jayapal R

External Viva

Name of Examiners Signature with date

2
 RV INSTITUTE OF TECHNOLOGY AND MANAGEMENT®
(Affiliated to Visvesvaraya Technological University, Belagavi & Approved by AICTE, NewDelhi)
Bengaluru-560076

DEPARTMENT OF
ELECTRONICS AND COMMUNICATION ENGINEERING

DECLARATION

We, AKSHAT GUPTA-1RF20EC003, MOHAMMED NADEEM-1RF20EC028,

SHIVAM KUMAR RAI-1RF20EC042 and VIKRANT RANA-1RF1EC053, the
students of sixth semester DEPARTMENT OF ELECTRONICS AND
COMMUNICATION ENGINEERING, hereby declare that the mini project titled
“RECOGNITION OF DIABETIC RETINOPATHY USING CNN
ALGORITHM” has been carried out by us and submitted in partial fulfillment for
the award of the degree of Bachelor of Engineering in ELECTRONICS AND
COMMUNICATION ENGINEERING. We do declare that this work is not
carried out by any other students for the award of a degree in any other branch.

Name: Signature:
Place: Bengaluru
1. AKSHAT GUPTA
Date:
2. MOHAMMED NADEEM

3. SHIVAM KUMAR RAI

4. VIKRANT RANA
 ACKNOWLEDGEMENT

The successful presentation of the PROJECT would be incomplete without the

mention of the people who made it possible and whose constant guidance crowned
oureffort with success.

We would like to thank our Project Guide, Dr. Kavitha N, Assistant Professor,
Department of Electronics and Communication Engineering, RV Institute of
Technology and Management, Bengaluru, for her constant guidance and inputs.

We thank Dr. Prashant P Patavardhan, Professor and Head of the Department of

Electronics and Communication Engineering, RV Institute of Technology and
Management, Bengaluru, for his encouragement.

We would like to extend our gratitude to Dr. Jayapal R, Principal, RV Institute of

Technology and Management, Bengaluru, for providing us an opportunity to work on
a project in this institution.

We would like to thank all the Teaching and Non-Teaching Staff for their
cooperation.

We would like to extend our gratitude to the MANAGEMENT, RV Institute of

Technology and Management, Bengaluru, for providing all the facilities to present the
Project.

Finally, we extend our heartfelt gratitude to our family for their encouragement and
support without which we wouldn’t have come so far. Moreover, we thank all our
friends for their invaluable support and cooperation.

AKSHAT GUPTA-1RF20EC003
MOHAMMED NADEEM-1RF20EC028
SHIVAM KUMAR RAI-1RF20EC042
VIKRANT RANA-1RF20EC053
 ABSTRACT

Diabetes is a chronic end organ disease that occurs when the pancreas does not secrete
enough insulin or the body is unable to process it properly. Over time, diabetes affects the
circulatory system, including that of the retina. Diabetic retinopathy (DR) is a medical
condition where the retina is damaged because fluid leaks from blood vessels into the
retina. Ophthalmologists recognize diabetic retinopathy based on features, such as blood
vessel area, exudes hemorrhages, microaneurysms and texture. Early detection and
treatment can limit the potential for significant vision loss from diabetic retinopathy.

CNN algorithms are widely used for the detection of diabetic retinopathy (DR), it helps in
analyzing fundus images in diagnosis. The proposed approach includes 3 stages:
Preprocessing, Feature extraction and Classification. The proposed method is evaluated
on a publicly available dataset. Image pre-processing is the steps taken to format images
before they are used by model training and inference. Image preprocessing may also
decrease model training time and increase model inference speed. Feature extraction is a
part of the dimensionality reduction process, in which, an initial set of the raw data is
divided and reduced to more manageable groups. The last step includes a classification
algorithm which classifies the images into 5 categories namely: No DR, Mild DR,
Moderate DR, Severe DR and Prolific DR.

The manual process of periodic DR diagnosis and detection for necessary treatment, is
time consuming and unreliable due to unavailability of resources and expert opinion.
Therefore, computerized diagnostic systems which use Convolutional Neural Network
(CNN) architectures, are proposed to learn DR patterns from fundus images and identify
the severity of the disease. We have successfully been able to achieve a maximum of
95.40% accuracy using the model that we have developed. Thus, we have minimized
erroneous outputs and maximized accuracy to provide better result analysis and help
medical workers choose the required treatment accordingly.

)i
 TABLE OF CONTENTS

Abstract………………………………………………………………….. i

List of Figures and Tables………………………………………………..ii

List of Abbreviations…………………………………………………….iii

1. Chapter 1: Introduction 1-7

1.1 Problem Statement…………………………………………………..1

1.2 Literature Survey....……………………………………………….....2

1.3 Motivation…………………………………………………………...6
1.4 Objectives…………………………………………………................6
1.5 Organization of the Report…………………………………………..7

2. Chapter 2: Theory and fundamentals 8-13

2.1 Diabetic Retinopathy……………………………………………..8-10
2.1.1 Introduction……………………………………………………8

2.1.2 Complication……………………………..................................8
2.1.3 Prevention……………………………………………………...9
2.2 Deep Learning - An overview…………………………………...10-13
2.2.1 Introduction to Deep Learning (DL)………….........................10
2.2.2 About Convolution Neural Network(CNN)………………......11

3. Chapter 3: Methodology and Design Flow 14-17

3.1 Methodology…..................................................................................14
3.2 Algorithm…………………………………………………………...14
3.3 Flow chart/ Block diagram…………………………………………15
3.4 Platform/Software Used……………………………………………16

4. Chapter 4: Implementation 18-28

4.1 CODE…………………………………………………………...18-25
4.1.1 Libraries……………………………………………………...18
 4.1.4 CNN Architecture Block……………………………………21

4.1.5 Training and Epochs………………………………………...23

4.1.6 Graphical Representation…………………………………...24
4.1.7 Testing………………………………………………………24
4.1.8 Accuracy test………………………………………………..25
4.1.9 Visualization………………………………………………...25
4.2 Frontend HTML page………………………………………….26-28

5. Chapter 5: Results, Discussion 29-37

5.1 Fitting the Convolution Neural Network and training…….....29-30
5.2 Graphical Outputs…………………………………………....31-33
5.2.1 Model Accuracy…………………………………………...32
5.2.2 Model Loss………………………………………………...33
5.3 Confusion Matrix………………………………………………..34
5.4 Prevention of stage of diabetic retinopathy………………….35-37

6. Chapter 6: Conclusion and Future Scope……………………….38

7. References……………………………………………………...39-40
 LIST OF FIGURES
FIGURE NO. DESCRIPTION PAGE NO.

Fig 1.1 Stages of Diabetic Retinopathy 1

Fig 2.1 Healthy Vs Diabetic Eye 9

Fig 2.2 Average Pooling 12

Fig 2.3 Convolution Layers 12

Fig 2.4 Zero Padding 13

Fig 2.5 Stride 13

Fig 3.1 Flowchart 15

Fig 4.1 Data Distribution 19

Fig 5.1: Results of accuracy and loss obtained after 30

all the epochs

Fig 5.2 Graph for Model accuracy of the ML model 32

for different iterations

Fig 5.3 Graph for Model loss of ML model for 33

different iterations

Fig 5.4 Confusion matrix 34

Fig. 5.5 Output image for mild DR prediction vs 35

original

Fig. 5.6 Output image for moderate DR prediction 35

vs original

Fig. 5.7 Output image for no DR prediction vs 36

original
)ii
 ABBREVIATIONS AND NOMENCLATURE

Term Description

Accuracy A metric used to measure model performance in terms of how

correctly it predicts data points or image feature against the
ground truth.

AI Artificial Intelligence; reference to the simulation of human

intelligence in machines enabled by computer programs

API Application Programming Interface

APTOS Asia Pacific Tele-Ophthalmology Society

CNN Convolutional Neural Network

Confusion An N x N matrix used for evaluating the performance of a

Matrix classification algorithm

DR Diabetic Retinopathy; An eye condition caused by diabetes

complication that the blood vessels of the light-sensitive
tissues (retina) which can lead to total blindness

Diagnosis identification of the nature of the disease through examination

of the symptoms

Epoch A period of time

Kaggle An online data science environment that is used for

practicing of Data Science technologies

Keras A Deep Learning API is written in python and running on

top of the TensorFlow platform.

TP True Positive

TN True Negative

FP False Positive

FN False Negative
ii)iii
Term Description

TensorFlow Is an end-to-end ecosystem for Machine Learning with

tools, libraries, and community resources.

Validation The accuracy obtained from using the validation dataset

Accuracy

RGB Red, Green and Blue

ReLu Rectified Linear Unit; An activation function which is

used to remove any linear component.

TP True Positive

TN True Negative

FP False Positive

FN False Negative

)iv
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

CHAPTER 1

INTRODUCTION

This section briefly describes the background, definition of the problem statement, motivation
behind choosing the domain. It also consists of the important problem formulations and
objectives along with the scope of the project.
Diabetic retinopathy is an eye disease caused by diabetes that can lead to loss of vision or even
complete blindness. Diabetic retinopathy accounts for 12% of all new cases of blindness in the
United States, and is the leading cause of blindness for people aged 20 to 64 years. If caught
early enough, progression to vision impairment can be slowed if not altogether stopped,
however, this is often difficult because symptoms may appear too late to provide effective
treatment. Diabetic retinopathy (DR) has been estimated to affect about 93 million people
globally, though only half are aware of it.

1.1 Problem Statement

Current Limitations: Existing methods for DR detection often rely on manual assessment by
trained specialists, resulting in time-consuming and subjective evaluations that hinder scalability
and accessibility of diagnosis.
Need for Automated Solutions: There is a critical need for automated and reliable techniques
that can efficiently screen and detect DR from retinal images, enabling early diagnosis and
intervention.

Fig 1.1: Stages of Diabetic Retinopathy

Department of Electronics and Communication Engineering 1 1

11
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

1.2 Literature Survey

This sub-section contains information related to CNN-based approaches:

Diabetic Retinopathy (DR) is caused as a result of Diabetes Mellitus which causes

development of various retinal abrasions in the human retina. These lesions cause
hindrance in vision and in severe cases, DR can lead to blindness. DR is observed
amongst 80% of patients who have been diagnosed from prolonged diabetes for a
period of 10–15 years. The manual process of periodic DR diagnosis and detection for
necessary treatment, is time consuming and unreliable due to unavailability of resources
and expert opinion. Therefore, computerized diagnostic systems which use Deep
Learning (DL) Convolutional Neural Network (CNN) architectures, are proposed to
learn DR patterns from fundus images and identify the severity of the disease. [1] paper
proposes a comprehensive model using 26 state-of-the-art DL networks to assess and
evaluate their performance, and which contribute for deep feature extraction and image
classification of DR fundus images. In the proposed model, ResNet50 has shown
highest overfitting in comparison to Inception V3, which has shown lowest overfitting
when trained using the Kaggle’s EyePACS fundus image dataset.

CNN, hybrid CNN with ResNet, hybrid CNN with DenseNet are used on a huge dataset
with around 3662 train images to automatically detect which stage DR has progressed.
Five DR stages, which are 0 (No DR), 1 (Mild DR), 2 (Moderate), 3 (Severe) and 4
(Proliferative DR) are processed in the proposed work. The patient’s eye images are fed
as input to the model. The proposed deep learning architectures like CNN, hybrid CNN
with ResNet, hybrid CNN with DenseNet 2.1 are used to extract the features of the eye
for effective classification. The models achieved an accuracy of 96.22%, 93.18% and
75.61% respectively. The [2] paper concludes with a comparative study of the CNN,
hybrid CNN with ResNet, hybrid CNN with DenseNet architectures that highlights
hybrid CNN with DenseNet as the perfect deep learning classification model for
automated DR detection.

Department of Electronics and Communication Engineering,2022-23. 2 1

12
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

To help patients with the early detection of diabetic retinopathy, paper [3],they
proposed a computer vision-based technique to analyze and predict diabetes from the
retinal input images. Image preprocessing, segmentation, and feature extraction steps
are applied. Convolutional neural networks (CNN) and Support Vector Machine (SVM)
are trained with diabetic and non-diabetic retinal images. Results show CNN reports
better accuracy in DR compared to SVM.

Ophthalmologists use optical coherence tomography (OCT) and fundus photography

for the purpose of assessing the retinal thickness, and structure, in addition to detecting
edema, hemorrhage, and scars. Deep learning models are mainly used to analyze OCT
or fundus images, extract unique features for each stage of DR and therefore classify
images and stage the disease. A deep Convolutional Neural Network (CNN) with 18
convolutional layers and 3 fully connected layers is proposed [4] to analyze fundus
images and automatically distinguish between controls (i.e. no DR), moderate DR (i.e.
a combination of mild and moderate Non Proliferative DR (NPDR)) and severe DR
(i.e. a group of severe NPDR, and Proliferative DR (PDR)) with a validation accuracy
of 88%-89%, a sensitivity of 87%-89%, a specificity of 94% -95%, and a Quadratic
Weighted Kappa Score of 0.91–0.92 when both 5-fold, and 10-fold cross validation
methods were used respectively.

At first, the disease grading is targeted using different classification models.

Classification models used in this study are ResNet50, VGG-16, and VGG19. Three
datasets generated during pre-processing along with original images were used to train
the classification models. [5] These models were pre-trained on the ImageNet dataset
having 1000 classes and were finetuned by adding more layers for our disease grading
task. The advantage of using a pre-trained model is to mitigate the effect of fewer
datasets. shows the additional fine-tuning layers added to the original architecture of
the models.ResNet-50 is a model given by Microsoft and has the quality feed output of
some layer directly to the input of some other layer by bypassing layers in This study
[6] compared the performance of the revised Resnet model with other common CNNs
models (Xception, AlexNet, VggNet-s, VggNet-16 and ResNet-50.

Department of Electronics and Communication Engineering 3 131

Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

In examining said models, the results alluded to an over-fitting phenomenon, and the
outcome of the work demonstrates that the performance of the revised ResNet-50
(Train accuracy: 0.8395 and Test accuracy: 0.7432) is better than other common CNNs
(that is, the revised structure of ResNet-50 could avoid the overfitting problem, decease
the loss value, and reduce the fluctuation problem).

To solve the classification problems, many different types of ResNets are used, with
different numbers of layers: specifically, 18, 34, 50, 101, and 152 layers [7]. The
current deep learning framework for detecting and grading DR is ResNet-50. However,
the disadvantages of ResNet-50 are overfitting and fluctuations in accuracy, which
affect its accuracy in detecting DR. This study proposes three strategies to improve the
performance of ResNet-50, as follows:
Adaptive learning rate in ResNet-50. 2) Regularization: Regularization can be
employed to minimize the overfitting of the training model. 3) Obtain suitable features
from conv5_block1_out and conv5_block2_out in ResNet-50.

The proposed [8] approach uses deep features of ResNet-50 along with Random Forest
as a classifier for the detection and grading of diabetic retinopathy. High-level features
obtained from the average pooling layer of trained ResNet-50 are fed to a random
forest classifier. The depth of the deep network plays a pivotal role in their
performance. With the increase in layers, the model gives better performance.
However, it has also been observed that the addition of layers may increase the error
rate. This is named as an issue of vanishing gradients. The residual neural network,
also known as ResNet, was introduced to address this problem. Residual Network uses
the skip connection to indiscriminately allow some input to the layer to incorporate the
flow of information and also to prevent its loss, hence, addressing the problem of
vanishing gradients (which also suppresses the generation of some noise). Suppressing
the noise means averaging the models, which keeps a balance between precision and
generalization. To achieve higher precision and an estimated level of traversal, the most
efficient way is to increase more labeled data. The structure of ResNet speeds up the
training of ultra-deep neural networks and increases the model’s accuracy on large
training data.

Department of Electronics and Communication Engineering 4 1

14
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

The resolution of the input image has a direct impact on the DR grading
performance[9]. Generally, ResNet50 is designed for images of 224×224 input
resolution (He et al., 2016). In ResNet-50, a convolution layer with 6 a kernel size of
7×7 and a stride of 2 followed by a max-pooling layer is applied to dramatically
down sample the input image first. Therefore, using images with very small input
resolution may lose key features for DR grading, such as tiny lesions. In contrast, a
network fed with large resolution images can extract more fine-grained and dense
features at the cost of a smaller receptive field and a higher computational cost. In
this work, a range of resolutions is evaluated to identify the trade-off. Applying data
augmentation during training can increase the distribution variability of the input
images to improve the generalization capacity and robustness of a model of interest.

After the training of the residual network with 20 epochs, the features can extracted
from their average global pooling layer. These features will be detailed and unique as
this model averaged out all the activations of the final convolution layer. Due to
parameter limitations, the global average pooling does not require optimization.
Moreover, owing to spatial translation, it is more robust to the input as it summarizes
spatial information. The dropout was set to 0.2 to reduce the overfitting [10].

Department of Electronics and Communication Engineering. 5 1

15
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

1.3 MOTIVATION

According to the World Health Organization, diabetic retinopathy is the leading cause of
blindness among working-age adults, affecting up to 35% of people with diabetes worldwide. In
2021, Approximately 537 million adults (20-79 years) are living with diabetes. The total
number of people living with diabetes is projected to rise to 643 million by 2030 and 783
million by 2045. 3 in 4 adults with diabetes live in low- and middle-income countries. One out
of two people suffering from diabetes has been diagnosed with some stage of DR. Detection of
DR symptoms in time can avert the vision impairment in majority of cases, however such
revelation is difficult with present tools and methods. There has been a need for comprehensive
and automated DR detection tools and methods.

By developing an accurate and reliable method for detecting diabetic retinopathy using deep
learning techniques, One can potentially make a significant impact on patient outcomes and
improve the quality of life for many people. Generally this is what happens, suppose you have
to show your eyes to ophthalmologist , eyes are dilated and then there is consultation with the
ophthalmologist and this entire process takes a lot of time for diagnosis and the scarcity of
ophthalmologist makes it harder and expensive. By our technique we can simplify the process
of detection and make the entire consultation more efficient and cost effective.

1.4 OBJECTIVES

• Develop a CNN algorithm that accurately detects and classifies diabetic

retinopathy from retinal images.

• Explore and optimize the performance of the CNN algorithm by experimenting with
different network architectures, hyperparameters, and data augmentation techniques.

• Assess the interpretability of the CNN algorithm by analyzing the learned features
and generating visual explanations for the algorithm's predictions.

• Develop a user-friendly software or application that integrates the CNN algorithm for easy
deployment and utilization by healthcare professionals.

Department of Electronics and Communication Engineering 6 2

16
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

1.5 Organization of the Report

Chapter 1 deals with the introduction to Diabetic retinopathy and briefly describes the
background, definition of the problem statement, motivation behind choosing the
domain. It also consists of the important problem formulations and objectives of the
project.

Chapter 2 deals with the Theory and fundamentals of the area related to the problem
statement, it covers the working of existing methods. This chapter discussed the
problems that are faced by the people in different places.

Chapter 3 include the methodology implemented in the proposed project. It contains the
steps that are followed by the algorithm.

Chapter 4 deals with the implementation of the design model and development of the
fronted of the project.

Chapter 5 deals with the results and discussion of the project. It’s been discussed about
the output that has been achieved using the developed CNN model. It contains detailed
graphical representation of the accuracy and losses.

Chapter 6 deals with the conclusion and future scope of the work carried out. The
conclusion is a critical section that provides a summary of the key findings and
outcomes of the project. the future scope section outlines potential areas for future
research or improvement based on the results and findings of the project.

Department of Electronics and Communication Engineering. 7 1

17
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

Chapter 2

Theory and Fundamentals

This chapter highlights all the basic fundamentals, theory and principles of the
project in brief, starting from the introduction to what Diabetic Retinopathy is and
what CNN are.
2.1 Diabetic Retinopathy
2.1.1 Introduction
Diabetic retinopathy is a diabetes complication that affects the eyes. It's caused by
damage to the blood vessels of the light-sensitive tissue at the back of the eye
(retina).At first, diabetic retinopathy might cause no symptoms or only mild vision
problems. But it can lead to blindness. The condition can develop in anyone who has
type 1 or type 2 diabetes. The longer you have diabetes and the less controlled your
blood sugar is, the more likely you are to develop this eye complication[1].
According to a 2018 American Eye-Q ® survey conducted by the AOA, nearly half
of Americans didn't know whether diabetic eye diseases have visible symptoms
(often which the early stages of diabetic retinopathy does not). The same survey
found that more than one-third of Americans didn't know a comprehensive eye exam
is the only way to determine if a person's diabetes will cause blindness[2].
Symptoms:
You might not have symptoms in the early stages of diabetic retinopathy. As the
condition progresses, you might develop:
• Spots or dark strings floating in your vision (floaters)
• Blurred vision
• Fluctuating vision
• Dark or empty areas in your vision
• Vision loss

2.1.2 Complications
Diabetic retinopathy involves the growth of abnormal blood vessels in the retina.
Complications can lead to serious vision problems such as:

Department of Electronics and Communication Engineering,2022-23. 8 1

18
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

• Vitreous hemorrhage: The new blood vessels may bleed into the clear, jelly like
substance that fills the center of your eye. If the amount of bleeding is small, you
might see only a few dark spots (floaters). In more-severe cases, blood can fill the
vitreous cavity and completely block your vision. Vitreous hemorrhage by itself
usually doesn't cause permanent vision loss. The blood often clears from the eye
within a few weeks or months. Unless your retina is damaged, your vision will
likely return to its previous clarity.

• Retinal detachment: The abnormal blood vessels associated with diabetic

retinopathy stimulate the growth of scar tissue, which can pull the retina away
from the back of the eye. This can cause spots floating in your vision, flashes of
light or severe vision loss.

• Glaucoma: New blood vessels can grow in the front part of your eye (iris) and
interfere with the normal flow of fluid out of the eye, causing pressure in the eye
to build. This pressure can damage the nerve that carries images from your eye to
your brain (optic nerve).

• Blindness: Diabetic retinopathy, macular edema, glaucoma or a combination of

these conditions can lead to complete vision loss, especially if the conditions are
poorly managed. Fig 2.1 shows the comparison between a healthy eye and a
diabetic eye.

Fig 2.1: Healthy Vs Diabetic Eye

2.1.3 Prevention

You can't always prevent diabetic retinopathy. However, regular eye exams, good
control of your blood sugar and blood pressure, and early intervention for vision
problems can help prevent severe vision loss.

Department of Electronics and Communication Engineering,2022-23. 9 1

19
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

If you have diabetes, reduce your risk of getting diabetic retinopathy by doing the
following:
• Manage your diabetes: Make healthy eating and physical activity part of your daily
routine. Try to get at least 150 minutes of moderate aerobic activity, such as
walking, each week. Take oral diabetes medications or insulin as directed.

• Monitor your blood sugar level: You might need to check and record your blood
sugar level several times a day, or more frequently if you're ill or under stress. Ask
your doctor how often you need to test your blood sugar.

• Ask your doctor about a glycosylated hemoglobin test: The glycosylated

hemoglobin test, or hemoglobin A1C test, reflects your average blood sugar level
for the two to three month period before the test. For most people with diabetes,
the A1C goal is to be under 7%.

• Keep your blood pressure and cholesterol under control: Eating healthy foods,
exercising regularly and losing excess weight can help. Sometimes medication is
needed, too.

• Pay attention to vision changes: Contact your eye doctor right away if your vision
suddenly changes or becomes blurry, spotty or hazy.

2.2 Deep Learning - An overview

2.2.1 Introduction to Deep Learning (DL)

Deep learning imitates how humans gain knowledge. Deep learning is part of data
science, which covers statistics and predictive modelling. Deep learning helps data
scientists acquire, analyze, and interpret massive amounts of data. Deep learning
automates predictive analytics. Deep learning algorithms are hierarchical, increasing
in complexity and abstraction. Iterate till output is accurate. Data processing layers
inspired the name deep.

Department of Electronics and Communication Engineering,2022-23. 10 1

110
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

Deep learning models can be made using several ways. Learning rate decay, transfer
learning, starting again, and dropout are approaches.

Deep learning models process information like the brain and can be used for
numerous tasks. Most image recognition, NLP, and speech recognition tools use deep
learning. Self-driving cars and language translation services are using these tools.

2.2.2 About Convolutional Neural Network (CNN)

In contrast to CNN, the performance of machine learning algorithms like K-nearest

neighbor, SVM, etc. reaches saturation after reaching a certain number.
Consequently, we intend to use CNN to advance. A convolutional neural network
(CNN) is a form of Diabetic retinopathy Detection and Diagnosis using Deep
Learning Algorithms artificial neural network that is specifically made to process
pixel input and is used in image recognition and processing. Convolution typically
means that it takes in an input signal and applies a filter over it, essentially multiplies
the input signal with the kernel to get the modified signal. Mathematically, a
convolution of two functions f and g is defined as shown in eq.2.1
𝑚
( 𝑓 ∗ 𝑔 ) ( ⅈ ) =∑ 𝑔 ( 𝑗 ) ⋅ 𝑓 𝑖 − 𝑗 +
𝑗 =1
( 𝑚
2 )
( 2.1)

A multilayer perceptron-like system that has been optimized for low processing
demands is used by a CNN. An input layer, an output layer, and a hidden layer with
several convolutional layers, pooling layers, fully connected layers, and normalizing
layers make up a CNN's layers. A system that is significantly more effective and
easier to train for image processing and natural language processing is produced by
the removal of restrictions and increase in efficiency for image processing[2,13].

Pooling

It is the step of down sampling the image’s features through summing up the
information. The operation is carried out through each channel and thus it only affects
the dimensions (n_H, n_W)and keeps n_C intact.

Department of Electronics and Communication Engineering,2022-23. 11 1

111
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

Given an image, we slide a filter, with no parameters to learn, following a certain

stride, and we apply a function on the selected elements.

The pooling filters used in the model are:

Average pooling: we average on the elements present on the filter.
Max pooling: given all the elements in the filter, we return the maximum, Fig 2.2

gives an illustration of an average pooling

Fig 2.2: Average Pooling

Convolutional layer:
A convolutional layer contains a set of filters whose parameters need to be learned.
The height and weight of the filters are smaller than those of the input volume. Each
filter is convolved with the input volume to compute an activation map made of
neurons. We apply convolutional products, using many filters this time, on the input
followed by an activation function ψ. Fig 2.3 illustrates each layer involved in
convolution filter.

Fig 2.3: Convolution Layers

Department of Electronics and Communication Engineering,2022-23. 12 1

1112
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

Padding:
The pixels on the corner of the image (2D matrix) are less used than the pixels in the
middle of the picture which means that the information from the edges is thrown away.
To solve this problem, we often add padding around the image in order to take the
pixels on the edges into account. In convention, we pad with zeros and denote with p
the padding parameter which represents the number of elements added on each of the
four sides of the image. The following picture illustrates the padding of a grayscale
image (2D matrix) where p=1,

Fig 2.4: Zero Padding

Stride:
The stride is the step taken in the convolutional product. A large stride allows to shrink
the size of the output and vice-versa. We denote s the stride parameter.
The Fig 2.5 illustrates a convolutional product (sum of element-wise element per block)
with s=1:

Fig 2.5: Stride

Department of Electronics and Communication Engineering 13 29

113
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

Chapter 3

Methodology and Design Flow

This chapter gives us an insight about the methodology followed in the project along
with the details of the basic requirements needed to start working on the project.

3.1 Methodology
Based on the inputs provided, we initially recognize the features of diabetic
retinopathy .Later steps involve algorithms from deep learning like Convolutional
Neural Networks (CNN) with a combination of Artificial Neural Networks to
develop an integrated network which extracts the input features and produces the
required output. This will help in classifying the data into i)NO DR ii) MILD
DR iii) MODERATE DR iv)SEVERE DR v)PROLIFERATIVE DR based
on the stages and making use of the data provided.

3.2 Algorithm
The general algorithm followed for executing the project is:
Step 1: Obtaining appropriate Data sets

Step 2: Pre-processing the Data sets for developing feature matrix

Step 3: Split the Data sets into Train and Test sets

Step 4: Develop multi-layered Neural network with the use of CNN.

Step 5: Feed the training set to the model for feature extraction

Step 6: Define the iterations or epochs accordingly to obtain a good accuracy.

Step 7: Once the model is trained with the training data set, expose it to the test data
Step 8: Obtain validation loss and accuracy, training loss and accuracy graphs.

Step 9: Generate heat and confusion matrix for the obtained results

Step 10: Give any test set input to the network to check the predictions and
classifications into five categories into i)NO DR ii) MILD DR iii) MODERATE
DR iv)SEVERE DR v)PROLIFERATIVE DR .
13
Department of Electronics and Communication Engineering 14
|
114
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

3.3 Flow chart/ Block diagram

Fig 3.1: Flowchart

Fig. 3.1 represents the design flow of our project in the form of fundus image The dataset
comprised of 3662 retinal fundus images. Out of which 2490 images are used for training, 439
images for validation and 733 for testing.

Pre-processing is a crucial process done before putting scanned fundus images into the deep
convolutional neural network. The model reads the image files (stored in data folder). After
decoding the PNG content it yields RGB grids of channels for the pixels. Convolution and
pooling layers are the standard layers in deep convolutional neural networks. Input feature maps
are convolved by convolution layers with kernel filters, which then generate output feature maps
with an optional non-linear function. The pooling layer only uses one value from the input map,
forcing the feature maps' resolution to drop so that the output feature maps can maintain their
local deformation invariance. Following the extraction of features from convolutional and
pooling layers, fully connected layers combine the output features into a complete feature vector
using maximum and average pooling operations.

Department of Electronics and Communication Engineering 15 14

115
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

Additionally, it is necessary to appropriately change the image size before feeding an

ultra-scanned image into a separate DCNN. So, we resize all the images in the dataset
into 256*256 pixels. These should be transformed into floating-point tensors for neural
network input. The extraction of features from scanned fundus images is the basics of
image analysis, which contributes to the accurate investigation of breast cancer in
patient.

Model fitting is a measure of how well a machine learning model generalizes to

similar data to that on which it was trained. A model that is well-fitted produces more
accurate outcomes. A model that is over fitted matches the data too closely. A model
that is under fitted doesn't match closely enough. After testing the results are analysed
in the form of confusion matrix and heat map.

3.4 PLATFORM/SOFTWARE USED

JUPYTER NOTEBOOK

Jupyter notebook lets anyone create and execute arbitrary Python code in the browser.
It's ideal for machine learning, data analysis, and education. A notebook is saved with
an .jpynb extension.

Python 3.10.0
Python libraries and frameworks offer a reliable environment which reduces software
development time significantly. Python is consistent, simple, flexible, platform
independent and has wide community which makes it most appropriate for machine
learning Python includes a modular machine learning library PyBrain, Tensorflow,
Keras, NumPy etc which offers many algorithms for machine learning task. Version

3.10.0 was used for this project.

VS CODE
Visual Studio Code, also commonly referred to as VS Code is a source-code editor
made by Microsoft with the Electron Framework, for Windows, Linux and macOS.
It is used for compiling the HTML code for frontend.

15
Department of Electronics and Communication Engineering 16
116
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

FLASK
Flask is a micro web framework written in python. It is classified as a
microframework because it does not require particular tools or libraries. It has no
database abstraction layer, form validation, or any other components where pre-
existing third-party libraries provide common functions. However, Flask supports
extensions that can add application features as if they were implemented in Flask
itself. Extensions exist for object relational framework, form validation, upload
handling, various open authentication technologies and several common framework
related tools.
Python Frontend framework that helps in converting the backend machine learning
model to web app that make the classification more interactive and gives a better user
experience.

15
Department of Electronics and Communication Engineering 17
117
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

Chapter 4

Design and Development

4.1 CODE:

4.1.1 Libraries

# Import the necessary packages

import pandas as pd
import numpy as np
import tensorflow as tf
from tensorflow import keras
import os
import matplotlib.pyplot as plt
import PIL
import seaborn as sns
import plotly
import plotly.graph_objs as go
from sklearn.model_selection import train_test_split
from sklearn.utils import shuffle
from plotly.offline import iplot, init_notebook_mode
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.applications.resnet50 import ResNet50
from tensorflow.keras.applications.inception_resnet_v2 import InceptionResNetV2
from tensorflow.keras.layers import *
from tensorflow.keras.models import Model, load_model
from tensorflow.keras.initializers import glorot_uniform
from tensorflow.keras.utils import plot_model
from IPython.display import display
from tensorflow.keras import backend as K
from tensorflow.keras.optimizers import SGD
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.models import Model, Sequential
from tensorflow.keras.callbacks import ReduceLROnPlateau, EarlyStopping,
ModelCheckpoint,LearningRateScheduler

Department of Electronics and Communication Engineering 18 1

118
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

4.1.2 Data Representation And Statistics

Fig 4.1: Dataset Distribution

Department of Electronics and Communication Engineering 19 1

119
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

4.1.3 Shuffling Data and Data Division

Department of Electronics and Communication Engineering 20 1

120
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

4.1.4 CNN Architectural Block

def res_block(X, filter, stage):

# Convolutional_block
X_copy = X

f1 , f2, f3 = filter

# Main Path
X = Conv2D(f1, (1,1),strides = (1,1), name ='res_'+str(stage)+'_conv_a',
kernel_initializer= glorot_uniform(seed = 0))(X)
X = MaxPool2D((2,2))(X)
X = BatchNormalization(axis =3, name = 'bn_'+str(stage)+'_conv_a')(X)
X = Activation('relu')(X)

X = Conv2D(f2, kernel_size = (3,3), strides =(1,1), padding = 'same', name

='res_'+str(stage)+'_conv_b', kernel_initializer= glorot_uniform(seed = 0))(X)
X = BatchNormalization(axis =3, name = 'bn_'+str(stage)+'_conv_b')(X)
X = Activation('relu')(X)

X = Conv2D(f3, kernel_size = (1,1), strides =(1,1),name ='res_'+str(stage)

+'_conv_c', kernel_initializer= glorot_uniform(seed = 0))(X)
X = BatchNormalization(axis =3, name = 'bn_'+str(stage)+'_conv_c')(X)

# Short path
X_copy = Conv2D(f3, kernel_size = (1,1), strides =(1,1),name ='res_'+str(stage)
+'_conv_copy', kernel_initializer= glorot_uniform(seed = 0))(X_copy)
X_copy = MaxPool2D((2,2))(X_copy)
X_copy = BatchNormalization(axis =3, name = 'bn_'+str(stage)+'_conv_copy')
(X_copy)

# ADD
X = Add()([X,X_copy])
X = Activation('relu')(X)

# Identity Block 1
X_copy = X

# Main Path
X = Conv2D(f1, (1,1),strides = (1,1), name ='res_'+str(stage)+'_identity_1_a',
kernel_initializer= glorot_uniform(seed = 0))(X)
X = BatchNormalization(axis =3, name = 'bn_'+str(stage)+'_identity_1_a')(X)
X = Activation('relu')(X)

Department of Electronics and Communication Engineering 21 1

121
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

X = Conv2D(f2, kernel_size = (3,3), strides =(1,1), padding = 'same', name

='res_'+str(stage)+'_identity_1_b', kernel_initializer= glorot_uniform(seed = 0))(X)
X = BatchNormalization(axis =3, name = 'bn_'+str(stage)+'_identity_1_b')(X)
X = Activation('relu')(X)

X = Conv2D(f3, kernel_size = (1,1), strides =(1,1),name ='res_'+str(stage)

+'_identity_1_c', kernel_initializer= glorot_uniform(seed = 0))(X)
X = BatchNormalization(axis =3, name = 'bn_'+str(stage)+'_identity_1_c')(X)

# ADD
X = Add()([X,X_copy])
X = Activation('relu')(X)

# Identity Block 2
X_copy = X

# Main Path
X = Conv2D(f1, (1,1),strides = (1,1), name ='res_'+str(stage)+'_identity_2_a',
kernel_initializer= glorot_uniform(seed = 0))(X)
X = BatchNormalization(axis =3, name = 'bn_'+str(stage)+'_identity_2_a')(X)
X = Activation('relu')(X)

X = Conv2D(f2, kernel_size = (3,3), strides =(1,1), padding = 'same', name

='res_'+str(stage)+'_identity_2_b', kernel_initializer= glorot_uniform(seed = 0))(X)
X = BatchNormalization(axis =3, name = 'bn_'+str(stage)+'_identity_2_b')(X)
X = Activation('relu')(X)

X = Conv2D(f3, kernel_size = (1,1), strides =(1,1),name ='res_'+str(stage)

+'_identity_2_c', kernel_initializer= glorot_uniform(seed = 0))(X)
X = BatchNormalization(axis =3, name = 'bn_'+str(stage)+'_identity_2_c')(X)

# ADD
X = Add()([X,X_copy])
X = Activation('relu')(X)

return X

input_shape = (256,256,3)

#Input tensor shape

X_input = Input(input_shape)

#Zero-padding

X = ZeroPadding2D((3,3))(X_input)

# 1 - stage

Department of Electronics and Communication Engineering 22 1

122
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

X = Conv2D(64, (7,7), strides= (2,2), name = 'conv1', kernel_initializer=

glorot_uniform(seed = 0))(X)
X = BatchNormalization(axis =3, name = 'bn_conv1')(X)
X = Activation('relu')(X)
X = MaxPooling2D((3,3), strides= (2,2))(X)

# 2- stage

X = res_block(X, filter= [64,64,256], stage= 2)

# 3- stage

X = res_block(X, filter= [128,128,512], stage= 3)

# 4- stage

X = res_block(X, filter= [256,256,1024], stage= 4)

# # 5- stage

#X = res_block(X, filter= [512,512,2048], stage= 5)

#Average Pooling

X = AveragePooling2D((2,2), name = 'Averagea_Pooling')(X)

#Final layer

X = Flatten()(X)
X = Dense(5, activation = 'softmax', name = 'Dense_final', kernel_initializer=
glorot_uniform(seed=0))(X)

model = Model( inputs= X_input, outputs = X, name = 'Resnet18')

model.summary()

4.1.5 Training and Epochs

Department of Electronics and Communication Engineering 23 1

123
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

4.1.6 Graphical Representation

#plot of training loss and validation loss

plt.plot(history.history['loss'])
plt.plot(history.history['val_loss'])
plt.title('Model Training and Validation loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(['train_loss','val_loss'], loc = 'upper right')
plt.show()

#plot of training accuracy and validation loss

plt.plot(history.history['accuracy'])
plt.plot(history.history['val_accuracy'])
plt.title('Model Training and Validation Accuracy')
plt.ylabel('accuracy')
plt.xlabel('epoch')
plt.legend(['train_accuracy','val_accuracy'], loc = 'lower right')
plt.show()

#Loading the trained weights

model.load_weights("retina_weights.hdf5")

4.1.7 Testing

# Assigning label names to the corresponding indexes

labels = {0: 'Mild', 1: 'Moderate', 2: 'No_DR', 3:'Proliferate_DR', 4: 'Severe'}

# Loading images and their predictions

from sklearn.metrics import confusion_matrix, classification_report, accuracy_score

# import cv2

prediction = []
image = []

#Enter filepath
testpic = "./train/Moderate/00e4ddff966a.png"

img = PIL.Image.open(testpic)
plt.imshow(img)
img = img.resize((256,256))
# converting image to array
img = np.asarray(img, dtype= np.float32)
# normalizing the image
img = img / 255

Department of Electronics and Communication Engineering 24 1

12
124
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

# reshaping the image in to a 4D array

img = img.reshape(-1,256,256,3)
# making prediction of the model
predict = model.predict(img)
# getting the index corresponding to the highest value in the prediction
predict = np.argmax(predict)
# appending the predicted class to the list
prediction.append(labels[predict])
print('Prediction= {}'.format(predict))
predicted_label = labels[predict]
print('Prediction Label:', predicted_label)

4.1.8 Accuracy test

4.1.9 Visualization

# Visualizing the results

import random
fig=plt.figure(figsize = (100,100))
for i in range(30):
j = random.randint(0,len(image))
fig.add_subplot(30, 1, i+1)
plt.xlabel("Prediction: " + prediction[j] +" Original: " + original[j])
plt.imshow(image[j])
fig.tight_layout()
plt.show()

Department of Electronics and Communication Engineering 25 1

125
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

4.2 Frontend HTML page

Fig 4.2: Webpage

Fig 4.3: Error case

Department of Electronics and Communication Engineering 26 1

126
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

Fig 4.4: Select Image

Fig 4.5: Upload File

Department of Electronics and Communication Engineering 27 1

127
 Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

Fig 4.6: Displaying Output

In the frontend application for the report on diabetic retinopathy classification, the user interface
(UI) is designed to provide a user-friendly experience. The UI is shown in Fig 4.2. It includes an
image upload feature that allows users to upload retinal images for classification as shown in
Fig 4.4. After the image is uploaded, the frontend displays the uploaded image, performs
classification using a trained ML model, and presents the classification result to the user. The
end output is displayed in Fig 4.6. The frontend also handles error cases, as shown in Fig 4.3,
gracefully, such as when no image is uploaded. The design focuses on simplicity, usability, and
delivering accurate classification results to assist in the diagnosis of diabetic retinopathy.

Department of Electronics and Communication Engineering 28 1

128
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

Chapter 5

Results and Discussions

5.1 Fitting the Convolutional Neural Network and training

Once the neural network is deployed and appropriately designed using activation
functions(Relu, softmax), weights, kernels and biases. The input retinal fundus images
which is split into train and test sets using the library specified below from
sklearn. from sklearn.model_selection import train_test_split

The dataset comprised of 3662 retinal fundus images.Out of which 2490 images are
used for training ,439 images for validation and 733 for testing. An epoch means
training the neural network with all the training data for one cycle. In an epoch, we
use all of the data exactly once. A forward pass and a backward pass together are
counted as one pass: An epoch is made up of one or more batches, where we use a
part of the dataset to train the neural network.

The Fig. 5.1 represents the number of iterations or Epochs which the training data was
passed through. A total of 32 batches of data were trained with 15 epochs running for
each batch and obtaining the accuracy of prediction and model loss/accuracy.

At the end of the compilation of all the epoch cycles, the values for the model
accuracy, model loss, validation accuracy and validation loss efficiencies are obtained
as depicted in the Table 5.1.

Table 5.1: Results of accuracy and loss obtained after all the epochs

Sl. Attributes Efficiency

No.
1 Training Accuracy 95.40%
2 Training Loss 0.2049
3 Validation Accuracy 88.13%
4 Validation Loss 0.6468

The last Epoch shows us the results at the end of 15 iterations. As we can see that the
result obtained is about 88.13%.

Department of Electronics and Communication Engineering 2923

129
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

Fig 5.1: Results of accuracy and loss obtained after each epoch

Department of Electronics and Communication Engineering 30 1

130
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

5.2 Graphical Outputs

i. Model Accuracy : Machine learning model accuracy is the measurement used

to determine which model is best at identifying relationships and pattern
between variables in a dataset based on the input, or training, data. The better a
model can generalize to ‘unseen’ data, the better predictions and insights it can
produce, which in turn deliver more business value.

ii. Model loss: The loss function in a neural network quantifies the difference
between the expected outcome and the outcome produced by the machine
learning model. From the loss function, we can derive the gradients which are
used to update the weights. The average over all losses constitutes the cost.
iii. Validation Accuracy: It is the sum of errors made for each example in training
or validation sets. Loss value implies how poorly or well a model behaves
after each iteration of optimization. An accuracy metric is used to
measure the algorithm's performance in an interpretable way.
iv. Validation Loss: "Validation loss" is the loss calculated on the validation set,
when the data is split to train / validation / test sets using cross-validation.
When comparing test accuracy to training accuracy, test accuracy refers to how
well the trained model recognises independent images that were not utilised in
training. A
Accuracy = Number of correct predictions/ Total number predictions.
Loss is measured with the help of the loss function.
The difference between the outcome produced by the machine learning model
and the intended outcome is measured by the loss function in a neural network.

25
Department of Electronics and Communication Engineering 31
|P
131
ag
e
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

5.2.1 Model Accuracy

Fig 5.2: Graph for Model accuracy of the ML model for different iterations

The model accuracy as depicted, has 2 main features, curve 1 which is for the training
set represented in blue and curve 2 which is the test set represented in yellow. As it can
be seen that both the curves increase exponentially which denotes that the accuracy of
our model increases periodically with every passing epoch. By comparing the model
predictions with the actual values in terms of a percentage, it determines how well our
model predicts. The graphs in Fig 5.2 shows the plot of training and validation
accuracy when the patience parameter was set to 5, 15, 25 respectively.

26
Department of Electronics and Communication Engineering 32
|P
132
ag
e
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

5.2.2 Model Loss

Fig 5.3: Graph for Model loss of ML model for different iterations

The model loss as depicted, has 2 main features, curve 1 which is for the training set
represented in blue and curve 2 which is the test set represented in yellow. As it can be
seen in Fig. 5.3 both the curves decrease with every passing epoch. This denotes that
the model and validation loss of our model decreases from time to time and hence
reduces erroneous outputs. The graphs in Fig 5.3 shows the plot of training and
validation loss when the patience parameter was set to 5, 15, 25 respectively.

27
Department of Electronics and Communication Engineering 33
|P
133
ag
e
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

5.3 Confusion Matrix

By displaying the true and false predictions for each class, the confusion matrix goes
beyond classification accuracy. A confusion matrix in the context of a binary
classification job is a 2x2 matrix. It is a 5x5 matrix if there are five separate classes,
and so on. We won't receive the TP, TN, FP, and FN values immediately in the multi-
class classification problem as we would in the binary classification problem. For each
class, a calculation is required.

FN: A class's False-negative value is the sum of the values in the relevant rows,
excluding the TP value. FP: A class's False-positive value is the total of all of the
values in the relevant column, excluding the TP value. TN: The total of all columns
and rows, excluding those for the class for which we are computing the values, will
represent the True Negative value for a given class. TP: When the actual value and the
anticipated value are the same, that is a true positive value.

Fig 5.4: Confusion matrix of the model

28
Department of Electronics and Communication Engineering 34
|P
134
ag
e
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

5.4 Prediction of stage of diabetic retinopathy

The figures shown in this section depicts the expected outputs of the test data set for
the five cases:0)mild DR 1)moderate DR 2)no DR 3) proliferative DR 4)severe DR.

Prediction: mild original: mild

Fig. 5.5: Output image for mild DR prediction vs original

The Fig. 5.5 shows the output for the test case image which is rightly predicted as mild
DR. Once the input image of the test set is passed to the output variable, the image is
processed through all the convolutional network layers and gives the correct
prediction at the end of the process. It also mentions the class 0 to which it belongs.

Prediction: moderate original: moderate

Fig. 5.6: Output image for moderate DR prediction vs original

The Fig. 5.6 shows the output for the test case image which is rightly predicted as
moderate DR. Once the input image of the test set is passed to the output variable, the
image is processed through all the convolutional network layers and gives the correct
prediction at the end of the process. It also mentions the class 1 to which it belongs.

30
Department of Electronics and Communication Engineering 35
|P
135
ag
e
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

Prediction: No DR original: No DR
Fig. 5.7: Output image for no DR prediction vs original

The Fig 5.7 shows the output for the test case image which is rightly prediction as no
DR. The class for the image is 3.

Prediction: proliferative original: proliferative

Fig. 5.8: Output image for proliferative DR prediction vs original

The Fig 5.8 shows the output for the test case image which is rightly predicted as
proliferative DR. The class for the image is 4.

31
Department of Electronics and Communication Engineering 36
|P
136
ag
e
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

Prediction: severe original: severe

Fig. 5.9: Output image for severe DR prediction vs original

The Fig 5.9 shows the output for the test case image which is rightly predicted as
severe DR. The class for the image is 5.

31
Department of Electronics and Communication Engineering 37
|P
137
ag
e
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

Chapter 6

CONCLUSION AND FUTURE SCOPE

Our project aimed at giving a suitable solution to aid the process of predicting the
correct stage of Diabetic Retinopathy by obtaining inputs from the fundus images.
Hence, we have come up with a user-friendly approach in which we have trained
our model in different supervised classification algorithms to improve its accuracy.

We have been able to increase our accuracy to 95.40% which is a very efficient
number to rely on. We have also optimized the number of epochs and training
cycles for our input fundus images which reduces the error and improves the
model’s overall performance. Our main objective for taking up this project has
been achieved which was of making a user-friendly interface where the user gives
the input fundus images and our model predicts the stage of the DR.

With the advent of India’s own electronic health record system (EHR). EHR will
facilitate automatic screening and diagnosis of diabetic retinopathy. This can assist
medical professionals in identifying at-risk individuals and initiating quick actions
that improve patient outcomes.

India's very own telemedicine program eSanjeevani will play a major role in our
project as. The created project may be connected with telemedicine platforms,
allowing for remote diabetic retinopathy screening and diagnosis. For patients in
rural and underdeveloped areas in particular, this may boost access to affordable
healthcare. By enhancing the model, it is possible to diagnose diabetic retinopathy
in real-time, which would increase screening effectiveness and reduce the burden
on healthcare providers. The project can be further extended to be able to directly
be applied to a fundus camera where after taking an image of the retina of the
patient , the algorithm can easily detect to which stage of DR it is. Thereby
reducing the time spent on diagnosis of a patient and result in more efficiency and
accuracy.

Department of Electronics and Communication Engineering 3823

138
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

REFERENCES

[1] Dolly Das, Saroj Kumar Biswas, and Sivaji Bandyopadhyay “ Detection of Diabetic
Retinopathy using Convolutional Neural Networks for Feature Extraction and
Classification (DRFEC), ”IEEE Trans Geosci Remote Sens, vol.15,no. 3, pp.18-21, 29
Nov 2022.

[2] Yasashvini R.,Vergin Raja Sarobin M. ,Rukmani Panjanathan, Graceline Jasmine

S.,Jani Anbarasi L.ORCID “ Diabetic Retinopathy Classification Using CNN and
Hybrid Deep Convolutional Neural Networks ,” 42nd Annual International
Conference of the IEEE Engineering in Medicine & Biology Society (EMBC),Marsellei,
France,16 September 2022,pp. 55-61.

[3] Jaichandran R, Vithiavathi Sivasubramanian, Jaya Prakash ,Varshni M “Detection of

Diabetic Retinopathy Using Convolutional Neural Networks,” 2022 IEEE International
Conference on Information Reuse and Integration (IRI). Kolkata, India,18 may 2022,pp.
402-411.

[4] Mohamed Shaban,Zeliha Ogur,Ali Mahmoud,Andrew Switala,Ahmed Shalaby,Hadil

Abu Khalifeh,Mohammed Ghazal,Luay Fraiwan,Guruprasad Giridharan,Harpal
Sandhu,Ayman S. El-Baz “A convolutional neural network for the screening and staging
of diabetic retinopathy,” 39th Annual International Conference of the IEEE Engineering
in Medicine and Biology Society (EMBC),Dubai, UAE,June 22, 2021,pp-124-131.

[5] Saleh Albahli, Ghulam Nabi Ahmad Hassan Yar “ Detection of Diabetic Retinopathy
Using Custom CNN to Segment the Lesions,” Int. J. Eng. Adv. Technol., vol. 8, no. 6,
pp. 100-104,25 November 2021.

[6] Chun-Ling Lin, Kun-Chi Wu “Development of revised ResNet-50 for diabetic

retinopathy detection,” PLoS One Journal, vol. 14, no. 6, pp. 1–11, 2023 Apr 19.

[7] Chun-Ling Lin ,Kun-Chi Wu “Development of Revised ResNet-50 for Diabetic

Retinopathy Detection,” Computer Methods Programs Biomed., vol.153 ,no.18, pp.
144-149,January 4th, 2023.

Department of Electronics and Communication Engineering,2022-23. 39 1

139
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

[8] Muhammad Kashif Yaqoob ,Syed Farooq Ali,Muhammad Bilal ,Muhammad

Shehzad Hanif ,Ubaid M. Al-Saggaf “ ResNet Based Deep Features and Random Forest
Classifier for Diabetic Retinopathy Detection, ” International Conference on
Computational Intelligence ,Brisbane, Australia, 4 June 2021,pp. 1480–1492.

[9] Yijin Huanga, Li Lina,c, Pujin Chenga , Junyan Lyua,d, Roger Tamb, Xiaoying
Tanga “ Identifying the key components in ResNet-50 for diabetic retinopathy grading
from fundus images: a systematic investigation, ” International Conference on Expert
System Application, Geneva, Switzerland,17 Oct 2022,pp 212-219.

[10] Rajesh K, Santhanam A, Sridhar M, Dr. J. Mohan “ Diabetic Retinopathy

Detection using Convolutional Neural Network, ” International Journal of Advanced
Research in Science, Communication and Technology (IJARSCT),Volume. 2,no. 3, pp.
34-41, May 2022.

[11] C. Stauffer and W. E. L. Grimson, “Adaptive background mixture models for real-
time tracking,” in Proceedings. 2021 IEEE Computer Society Conference on Computer
Vision and Pattern Recognition, Collins, Colorado,USA, June 2021, pp. 246–252.

[12] O. Perdomo, S. Otalora, F. Rodríguez, J. Arevalo, and F. A. González, “A novel

machine learning model based on exudate localization to detect diabetic
macularedema,” In Proceedings of the Ophthalmic Medical Image Analysis Third
International Confernce, Athens, Greece, October 2021,pp. 120-128.

[13] Yashal Shakti Kanungo, Bhargav Srinivasan, Dr. Savita Choudhary, “Detecting
Diabetic Retinopathy using Deep Learning” International Conference on Expert System
Application, Chennai, India, Sept 2021,pp.260-267.

[14] Mahmut Karakaya* and Recep E. Hacisoftaoglu, “Comparison of smartphone-

based retinal imaging systems for diabetic retinopathy detection using deep learning”
24th National and 2nd International African Conference on Biomedical Engineering
(ICBME) ,Johannesburg, South Africa, 28 March 2022,pp. 1-6.
Department of Electronics and Communication 40 1
140
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

[15] R.Bhargavi, R. Geetha Nalini, K.Poorna Subhash, K. Sainadh, “Diagnosis of Diabetic

Retinopathy Using Transfer Learning, 40th International conference on Biomedical
application, Texas, USA, April 2022,pp.145-153.

[16] D. Y. Carson Lam, “Automated detection of diabetic retinopathy using deep learning,”
AMIA Summits on Translational Science Proceedings, vol. 13,no. 6,pp 100-105, Feb 2020.

[17] F. Zabihollahy, A. Lochbihler, and E. Ukwatta, “Deep learning based approach for
fully automated detection and segmentation of hard exudate from retinal images,”
Proceedings of the Medical Imaging 2019: Biomedical Applications in Molecular,
Structural, and Functional Imaging vol.2020, Springer, San Diego, CA, USA, January
2020,pp. 403-408.

[18] Kumar, S. and Kumar, B., 2020, February. Diabetic Retinopathy Detection by
Extracting Area and Number of Microaneurysm from Colour Fundus Image. In 2020 5th
International Conference on Signal Processing and Integrated Networks (SPIN),San
Francisco, USA, 10 Jan 2020,pp. 359-364.

[19] Asiri, N., Hussain, M. and Abualsamh, H.A., 2022. Deep Learning based Computer-
Aided Diagnosis Systems for Diabetic Retinopathy in 25th International conference on
Medical electronics,Putrajaya,Malaysia,19 April 2022 , pp. 67-71.

[20] Gulshan, V., Peng, L., Coram, M., Stumpe, M.C., Wu, D., Narayanaswamy, A.,
Venugopalan, S., Widner, K., Madams, T., Cuadros, J. and Kim, R., Development and
validation of a deep learning algorithm for detection of diabetic retinopathy in retinal
fundus photographs. Jama, 316(22),pp.2402-2410, 2022.

[21] Mohammadian, S., Karsaz, A. and Roshan, Y.M., 2020, November. Comparative
Study of Fine-Tuning of Pre-Trained Convolutional Neural Networks for Diabetic
Retinopathy Screening. In 2020 24th National and 2nd International Iranian Conference
on Biomedical Engineering (ICBME),Tehran, Iran,12 Feb 2020, pp. 1-6.

Department of Electronics and Communication 41 1

140
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

[22] Wan, S., Liang, Y. and Zhang, Y., 2018. Deep convolutional neural networks for
diabetic retinopathy detection by image classification. Computers & Electrical
Engineering journal, vol. 24, pp.274-282, 17 July 2018

[23] Mansour, R.F., 2018. Deep-learning-based automatic computer-aided diagnosis

system diabetic retinopathy. Biomedical engineering conference,Munich,Germany,20
sept 2018,pp.41-57.

[24] Dutta, S., Manideep, B.C., Basha, S.M., Caytiles, R.D. and Iyengar, N.C.S., 2021.
Classification of Diabetic Retinopathy Images by Using Deep Learning Models.
International Journal of Grid and Distributed Computing, vol. 11,Nov 2021, pp.89-
106.

[25] Gao, Z., Li, J., Guo, J., Chen, Y., Yi, Z. and Zhong, J.,“2020. Diagnosis of
Diabetic Retinopathy Using Deep Neural Networks”, Proc. 7th Int. Conf. Bio Signals
Images Instrum. (ICBSII),Beijing, China,12 Mar 2021, pp. 1-4 .

[26] B. Harangi and A. Hajdu, “Detection of exudates in fundus images using a

Markovian segmentation model,” in 2014 36th Annual International Conference of the
IEEE Engineering in Medicine and Biology Society, IEEE, Chicago, IL, USA,
November 2022, pp. 130–133.

[27] C. Stauffer and W. E. L. Grimson, “Adaptive background mixture models for real-
time tracking,” in 2021 IEEE Computer Society Conference on Computer Vision and
Pattern Recognition, IEEE, Collins, Colorado, June 2021, , pp. 246–252.

Department of Electronics and Communication 42 1

140
Detection of Diabetic retinopathy using CNN algorithm R V Institute of Technology and Management®

Department of Electronics and Communication Engineering,2022-23. 1