VISVESVARAYA TECHNOLOGICAL UNIVERSITY

Jnana Sangama, Belagavi - 590018

Project Report on

“GLAUCOMA DETECTION USING MACHINE LEARNING

WITH OCT”
Submitted in partial fulfillment of the requirements for the award of the degree of

BACHELOR OF ENGINEERING

in

INFORMATION SCIENCE & ENGINEERING

By

MAYUR S KARKERA 4MT20IS020

Under the Guidance of

Mr. Sharath Kumar
Assistant Professor, ISE

DEPARTMENT OF INFORMATION SCIENCE & ENGINEERING

(Accredited by NBA)

MANGALORE INSTITUTE OF TECHNOLOGY & ENGINEERING

(A Unit of Rajalaxmi Education Trust®, Mangalore)
Autonomous Institute affiliated to VTU, Belagavi, Approved by AICTE, New Delhi
Accredited by NAAC with A+ Grade & ISO 9001:2015 Certified Institution
Badaga Mijar, Moodabidri-574225, Karnataka
2023-2024
MANGALORE INSTITUTE OF TECHNOLOGY & ENGINEERING
(A Unit of Rajalaxmi Education Trust®, Mangalore)
Autonomous Institute affiliated to VTU, Belagavi, Approved by AICTE, New Delhi
Accredited by NAAC with A+ Grade & ISO 9001:2015 Certified Institution
Badaga Mijar, Moodabidri-574225, Karnataka

DEPARTMENT OF INFORMATION SCIENCE & ENGINEERING

(Accredited by NBA)

CERTIFICATE

This is to certify that the project work entitled “GLAUCOMA DETECTION USING MACHINE

LEARNING WITH OCT” is a bonafide work carried out by MAYUR S KARKERA

(4MT20IS020) in partial fulfillment for the award of degree of Bachelor of Engineering in
Information Science and Engineering of the Visvesvaraya Technological University, Belagavi
during the year 2023-24. It is certified that all corrections and suggestions indicated for Internal
Assessment have been incorporated in the report deposited in the departmental library. The project
has been approved as it satisfies the academic requirements in respect of project work prescribed for
the Bachelor of Engineering degree.

Mr. Sharath Kumar Prof. Manjunath H Dr. Prashanth C M

Project Guide Head of the Department Principal

External Viva
Name of the Examiners Signature with Date

1)

2)
 MANGALORE INSTITUTE OF TECHNOLOGY & ENGINEERING
Accredited by NAAC with A+ Grade, An ISO 9001: 2015 Certified Institution
(A Unit of Rajalaxmi Education Trust)
Affiliated to Visvesvaraya Technological University, Belagavi
Badaga Mijar, Moodabidri – 574 225 , Karnataka

DEPARTMENT OF INFORMATION SCIENCE & ENGINEERING

(Accredited by NBA)

DECLARATION
I MAYUR S KARKERA (4MT20IS020) student of 8th semester BE in Information Science &
Engineering, Mangalore Institute of Technology and Engineering, Moodabidri, hereby declare
that the project work entitled “GLAUCOMA DETECTION USING MACHINE LEARNING
WITH OCT”, submitted to the Visvesvaraya Technological University, Belagavi during the
academic year 2023-24, is a record of an original work done by us under the guidance of Mr. Sharath
Kumar, Assistant Professor, Department of Information Science & Engineering , Mangalore
Institute of Technology and Engineering Moodabidri. This project work is submitted in partial
fulfillment of the requirements for the award of the degree of Bachelor of Engineering in Information
Science & Engineering. The results embodied in this report have not been submitted to any other
University or Institute for the award of any degree.

Date: MAYUR S KARKERA

Place: Moodabidri
 ABSTRACT

This project is dedicated to advancing the diagnosis of glaucoma and its primary sub-conditions:
Choroidal Neovascularization (CNV), Diabetic Macular Edema (DME), and Drusen, recognizing
their significant impact on vision loss worldwide. Glaucoma, often termed the "silent thief of sight,"
leads to progressive optic nerve damage and irreversible vision loss. While elevated intraocular
pressure (IOP) stands as a primary cause, not all cases stem from high IOP, emphasizing the
multifactorial nature of its development, involving genetics, age, race, and family history. This
necessitates a comprehensive approach for early detection and management to mitigate the disease's
silent progression. To address the urgent need for early intervention, this project employs cutting-
edge machine learning and image analysis techniques, complementing traditional diagnostic
methods like tonometry and ophthalmoscopy. By leveraging these advanced technologies, the
project aims to enhance early detection capabilities, allowing for tailored approaches to diagnose
and manage glaucoma and its sub conditions. Early identification is critical due to the insidious
nature of glaucoma, often advancing unnoticed until irreversible damage occurs. Prioritizing early
intervention not only slows disease progression but also safeguards the quality of life for affected
individuals, emphasizing the importance of personalized diagnostic and treatment strategies for
various complications and subtypes of glaucoma

i
 ACKNOWLEDGEMENT
The satisfaction and the successful completion of this project would be incomplete without the
mention of the people who made it possible, whose constant guidance encouragement crowned my
efforts with success.

This project is made under the guidance of Mr. Sharath Kumar, Assistant Professor, Department
of Information Science & Engineering. I would like to express my sincere gratitude to my guide
for all the helping hand and guidance in this project.

I would like to thank my project coordinator Mr. Narendra UP, Associate Professor,
Department of Information Science & Engineering for his cordial support, valuable information
and guidance, which helped me in completing this project through the various stages.

I would like to thank our Principal Dr. Prashanth C M, for encouraging me and giving me an
opportunity to accomplish the project.

I also thank our management who helped me directly or indirectly in the completion of this project.

My special thanks to faculty members and others for their constant help and support.

Above all, I extend my sincere gratitude to our parents and friends for their constant
encouragement with moral support

MAYUR S KARKERA

ii
 TABLE OF CONTENTS

CONTENTS PAGE
NO.
ABSTRACT i
ACKNOWLEDGEMENT ii
TABLE OF CONTENTS iii
LIST OF TABLES vi
LIST OF FIGURES vii

CHAPTER NO. TITLE PAGE NO.

1 INTRODUCTION 1
1.1 Introduction 1
1.2 Problem Statement 1
1.3 Objectives (Purpose of the project) 2
1.4 Scope of the project 2
1.6 Organization of the report 2
2 LITERATURE SURVEY 4
2.1 Existing System 5
2.2 Limitations of existing systems 6
2.3 Proposed System 6
3 SYSTEM REQUIREMENTS SPECIFICATION 7
3.1 Overall description 7
3.1.1 Product Perspective 7
3.1.2 Product Functions 8
3.1.3 User classes and Characteristics 8
3.1.4 Design and Implementation constraints 9
3.1.5 Assumptions and Dependencies 9

iii
 3.2 Specific Requirements 9
3.2.1 Hardware Requirements 9
3.2.2 Software Requirements 10
3.2.3 Functional Requirements 12
3.2.4 Non-functional Requirements 13
4 GANTT CHART 15
5 SYSTEM DESIGN 16
5.1 Architectural Diagram 16
5.2 Use Case Diagram and Description 17
5.3 Sequence Diagram 17
5.4 Activity Diagram 18
5.5 Data flow Diagram 19
6 IMPLEMENTATION 21
6.1 System implementation 21
6.2 Workflow 22
7 TESTING 25
7.1 Introduction 25
7.2 Testing Approaches 25
7.3 Test Levels 26
7.4 Evaluation Metrics 26
7.5 Test Cases 27
8 RESULTS AND SNAPSHOTS 29
9 CONCLUSION AND FUTURE WORK 32
9.1 Conclusion 32
9.2 Future Work 33

iv
REFERENCES 34
PAPERS PRESENTED 35

v
 LIST OF TABLES

TABLE NO. TABLES PAGE

NO.

Table 4.1 Gantt chart of planning and scheduling of project 15

vi
 LIST OF FIGURES

FIG NO. TITLES PAGE NO.

Figure 4.1 Gantt Chart 15

Figure 5.1 Architecture of the system 16

Figure 5.2 Use Case design of the System 17

Figure 5.3 Sequence Diagram of the System 18

Figure 5.4 Activity Diagram of the System 19

Figure 5.5 Data Flow Diagram of the System 20

Figure 8.1 Home Page 29

Figure 8.2 About Us Page 29

Figure 8.3 Blog Page 30

Figure 8.4 Classifier Page 30

Figure 8.5 Contact Page 31

vii
 Chapter – 1
INTRODUCTION
 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

CHAPTER 1

INTRODUCTION

In this chapter, we are going to give introduction about our project titled as Lightning
Prediction and Alert System using deep learning models .

1.1 Introduction

Glaucoma, often referred to as the "silent thief of sight," is a prevalent and potentially
devastating eye disease characterized by the progressive damage to the optic nerve, leading to
irreversible vision loss. It is estimated that over 80 million individuals worldwide are affected
by this condition, making it one of the leading causes of blindness. While glaucoma primarily
affects the elderly population, it can also manifest in individuals of all ages, making its early
detection and management of paramount importance. This project is dedicated to the
development and enhancement of diagnostic methods for glaucoma, with a specific focus on
the identification of its primary sub-conditions: Choroidal Neovascularization (CNV),
Diabetic Macular Edema (DME), and Drusen. To fully comprehend the significance of these
sub-conditions and the urgency for their early detection, we must delve into the causes and
symptoms of glaucoma itself. he primary cause of glaucoma is elevated intraocular pressure
(IOP), a condition where the fluid within the eye, known as aqueous humor, is not adequately
drained. This leads to increased pressure within the eye, which, over time, damages the optic
nerve. Elevated IOP is a major risk factor, but it is essential to recognize that not all individuals
with glaucoma have high IOP, and not all individuals with high IOP develop glaucoma

1.2 Problem Statement

In response to the challenges faced in accurately detecting glaucoma using conventional

methods, we present a solution harnessing Machine Learning (ML) in conjunction with
Optical Coherence Tomography (OCT) technology. Traditional approaches to glaucoma
diagnosis often lack precision and consistency, leading to potential errors. Our initiative seeks
to address these shortcomings by leveraging ML algorithms to analyze OCT scans. By
extracting relevant features and patterns indicative of glaucomatous damage, our ML model
aims to provide a reliable and objective method for detecting glaucoma at an early stage.

Page ISE,1MITE, Moodabidri

Dept.1ofPage Page 1
 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

1.3 Objectives

The main objectives are as follows:

● Develop and enhance diagnostic methods for glaucoma, with a specific focus on the
identification of its primary sub-conditions: CNV, DME, and Drusen.
● Employ state-of-the-art machine learning and image analysis techniques to improve the
early detection of glaucoma and its subconditions.
● Advance our understanding and diagnostic capabilities to contribute to early intervention,
and the preservation of vision.

1.4 Scope

The scope of this project involves the utilization of machine learning (ML) algorithms in
conjunction with Optical Coherence Tomography (OCT) technology to innovate glaucoma
detection methods. By employing ML and deep learning techniques, including convolutional
neural networks (CNNs) and Random Forest algorithms, the project aims to enhance the
accuracy and efficiency of glaucoma diagnosis. Through the analysis of OCT images and
extraction of relevant features indicative of glaucomatous damage, the AI system seeks to
provide clinicians with precise and timely identification of the disease. Additionally, the
project emphasizes the integration of ML models into user-friendly interfaces accessible via
web and mobile platforms, enabling seamless analysis of OCT scans for real-time detection
and monitoring of glaucoma progression. Collaboration with ophthalmologists and
researchers ensures the validation and refinement of the AI models, fostering responsible
and effective utilization of technology for improved patient care and management of
glaucoma.

1.5 Organization of the Report

Chapter 1 Starts by introducing the topic, which gives an insight and in-depth analysis of
the project work, which was closely followed by the problem statement which tells about the
problems faced by users, followed by the objectives of the topic, which gave rise to the
significance of the topic, followed by the scope and crowned by the organization of
theproject.

Dept. of ISE, MITE, Moodabidri Page 2

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

Chapter 2 Deals with the literature review, Existing system, Limitations of existing system
and proposed system.

Chapter 3 Focuses on System requirements specification, Overall description, Specific

requirements like hardware and software requirements, functional and non-functional
requirements.

Chapter 4 Has Gantt chart which gives overall schedules of project.

Chapter 5 Which has Architectural diagram, Class diagram, Use case diagram, Sequence
diagram, Dataflow diagram.

Chapter 6 Focuses on the implementation part.

Chapter 7 Consists of results and snapshots of overall project.

Chapter 8 Is the conclusion part followed by future work.

Dept. of ISE, MITE, Moodabidri Page 3

 Chapter – 2
LITERATURE SURVEY
 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

CHAPTER 2

LITERATURE REVIEW

A literature survey or a literature review in a project report is that section which shows the
various analysis and research made in the field of our interest and the results already published,
taking into account the various parameters and the extent of the project.

In a paper by Jie Wang, In this paper researchers have made significant progress in developing
a deep learning algorithm for diagnosing and segmenting choroidal neovascularization (CNV)
in OCTA images. This novel algorithm has demonstrated exceptional accuracy in both tasks,
paving the way for its potential clinical application and improving the diagnosis and treatment
of CNV. By automating the process of CNV identification and segmentation, this deep learning
method can significantly reduce the workload of ophthalmologists and enhance the quality of
patient care. Additionally, the ability of this algorithm to accurately quantify CNV lesions
provides valuable insights for treatment planning and monitoring disease progression, further
optimizing patient outcomes. While further research is warranted to refine the algorithm and
evaluate its performance in real-world clinical settings, this deep learning approach holds
immense promise for revolutionizing CNV diagnosis and management.

In a paper by Filippo Arcado,This paper is proposed to develop deep learning (DL) models for
the automatic detection of optical coherence tomography (OCT) measures of diabetic macular
thickening (MT) from color fundus photographs (CFPs). Retrospective analysis on 17,997
CFPs and their associated OCT measurements from the phase 3 RIDE/RISE diabetic macular
edema (DME) studies. DL with transfer learning cascade was applied on CFPs to predict time-
domain OCT (TD-OCT)– equivalent measures of MT, including central subfield thickness
(CST) and central foveal thickness (CFT). MT was defined by using two OCT cutoff points:250
μm and 400 μm. A DL regression model was developed to directly quantify the actual CFT and
CST from CFPs. DL is capable of predicting key quantitative TD-OCT measurements related
to MT from CFPs. The DL models presented here could enhance the efficiency of DME
diagnosis in teleophthalmology programs, promoting better visual outcomes. Future research is
needed to validate DL algorithms for MT in the real-world.

Dept. of ISE, MITE, Moodabidri Page 4

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT
In a paper by Quang T.M.Pham, we propose a multi-scale deep learning model for drusen
segmentation. While most of the previous work on medical segmentation used UNet-based
architecture on the whole image , our method applies a multi scale learning method to make a
fine segmentation prediction. It is suitable for drusen segmentation with high-resolution fundus
images. To solve the high-resolution image problem, other researches tried to analyze on
cropped image , which may lose the global information. By combining global and local
information, our model is able to predict more accurate drusen segmentation masks. Moreover,
by exploiting the pre-trained model and the combination of different loss functions, we can
improve the performance of detecting drusen in the early stage of AMD. The experimental
results show that our method obtains the best Dice score on both our validation dataset and the
STARE dataset.This paper provides a fundamental step for further applications of deep learning
methods in the screening and diagnosis of AMD.

In a paper by Rakesh Patel,Age-related macular degeneration (AMD), diabetic retinopathy

(DR), drusen, choroidal neovascularization (CNV), and diabetic macular edema (DME) are
significant causes of visual impairment globally. Optical coherence tomography (OCT)
imaging has emerged as a valuable diagnostic tool for these ocular conditions. However,
subjective interpretation and inter-observer variability highlight the need for standardized
diagnostic approaches. The study concludes that the developed deep learning model using AI
techniques is highly effective in the automated detection of drusen, CNV, and DME in OCT
images. Further validation studies and research efforts are necessary to evaluate the
generalizability and integration of the model into clinical practice. Collaboration between
clinicians, policymakers, and researchers is essential for advancing diagnostic tools and
management strategies for AMD and DR. Integrating this technology into clinical workflows
can positively impact patient care, particularly in settings with limited access to
ophthalmologists.Future research should focus on collecting independent datasets, addressing
potential biases, and assessing real-world effectiveness. Overall, the use of machine learning
algorithms in conjunction with OCT imaging holds great potential for improving the detection
and management of drusen, CNV, and DME, leading to enhanced patient outcomes and vision
preservation.

2.1 Existing System

While previous efforts have explored automated glaucoma detection using OCT images,
achieving accurate classification of glaucoma subtypes remains a challenge. This project builds
upon these advancements by incorporating a transfer learning approach. We'll utilize a pre-
trained Convolutional Neural Network (CNN) called VGG-19. This powerful model, trained
Dept. of ISE, MITE, Moodabidri Page 5
 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

on a vast dataset of general images, possesses valuable image recognition skills. By leveraging
VGG-19's pre-trained knowledge as a foundation, we aim to develop a more efficient system.
During "fine-tuning," we'll adjust VGG-19's weights to focus on recognizing specific features
within OCT scans that differentiate drusen, CNV, DME, and normal retinas.It is also important
to acknowledge the limitations associated with transfer learning. The pre-trained model might
have inherent biases based on the original dataset (ImageNet), potentially requiring extensive
fine-tuning with a large and diverse collection of OCT images to achieve optimal performance
for glaucoma subclassification

2.2 Limitations of existing system

Existing VGG-19-based systems for glaucoma subtyping suffer from biases inherited from
general image datasets (like ImageNet) and limited generalizability to medical image analysis.
Imagine training a detective on fashion photos to find criminals - they might miss key details.
Similarly, VGG-19 might miss crucial features in OCT scans. Additionally, its general object
recognition skills might not translate well to the subtle variations in OCT scans needed for
glaucoma subtyping.

2.3 Proposed System

Building upon advancements in automated glaucoma detection with OCT images, this project
seeks a more accurate solution for classifying glaucoma subtypes. While existing systems
utilize transfer learning with models like VGG-19, these approaches can be limited by biases
inherent in the pre-trained datasets used for general image recognition. To address this, we
propose a novel approach using YOLOv8, a state-of-the-art object detection model built on the
CNN architecture. Unlike VGG-19, YOLOv8 is specifically designed for object detection tasks,
making it a more natural fit for identifying crucial features within OCT scans, such as the optic
nerve and retinal layers. Additionally, by training YOLOv8 from scratch on a specifically
curated dataset of OCT images labeled for glaucoma, CNV, DME, Drusen, and normal retinas,
we aim to minimize potential biases and achieve superior accuracy in classifying these
glaucoma subtypes. This targeted approach using YOLOv8 has the potential to significantly
improve early detection and subclassification compared to previous methods that rely on
transfer learning.

Dept. of ISE, MITE, Moodabidri Page 6

 Chapter – 3
SYSTEM REQUIREMENT
AND SPECIFICATION
 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

CHAPTER 3

SYSTEM REQUIREMENTS SPECIFICATION

A software requirements specification (SRS) is a description of a software system to be

developed. The software requirements specification lays out functional and non-functional
requirements, and it may include a set of use cases that describe user interactions that the
software must provide. Use cases are also known as functional requirements. In addition to
use cases the SRS also contains non-functional requirements. Non-functional requirements are
requirements which impose constraints on the design or implementation. For the hardware
requirements the SRS specifies the logical characteristics of each interface between the software
product and the hardware components. It specifies the hardware requirementslike memory
restrictions, cache size, the processor, RAM size etc. those are required for the software to run.
Software requirements specification is a rigorous assessment of requirements before the more
specific system design stages, and its goal is to reduce later redesign.

3.1 Overall Description

This document outlines the functionalities and considerations for a machine learning system
designed to detect glaucoma using Optical Coherence Tomography (OCT) images. OCT is a
non-invasive imaging technique that captures detailed cross-sectional images of the retina. This
system will utilize machine learning algorithms to analyze these OCT images and identify
potential signs of glaucoma, a leading cause of blindness.

3.1.1 Product Perspective

Ophthalmologists and healthcare professionals can leverage this system to:

● Improve diagnostic accuracy: Machine learning can assist in identifying subtle glaucomatous

features in OCT scans, potentially leading to earlier and more accurate diagnoses.

● Enhance efficiency: Automated analysis of OCT images can streamline the diagnostic

workflow, allowing doctors to dedicate more time to patient care and treatment planning.

● Support objective assessment: Machine learning models can provide a more consistent and

objective analysis of OCT scans compared to traditional subjective methods.

Dept. of ISE, MITE, Moodabidri Page 7

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

3.1.2 Product Function

The core functionality of this system is the automated detection of glaucoma in OCT images

using machine learning. This involves the following key steps:

● Data Preprocessing: OCT images will be preprocessed to ensure consistency in format

and data type. This may involve resizing, normalization, and other image manipulation

techniques.

● Feature Extraction: Machine learning algorithms will extract relevant features from the

preprocessed OCT images. These features could include measurements of retinal nerve fiber

layer thickness, cup-to-disk ratio, and other parameters indicative of glaucomatous changes.

● Model Training: The system will be trained on a large dataset of labeled OCT images.

Each image will be labeled as either positive (glaucoma present) or negative (glaucoma absent).

This training process allows the machine learning model to learn the complex relationships

between image features and the presence of glaucoma.

● Image Classification: Once trained, the system can be used to analyze new, unseen OCT

images. The model will classify each image as either positive or negative for glaucoma based

on the extracted features.

3.1.3 User classes and Characteristics

This system is primarily designed for:

● Ophthalmologists: This system can be a valuable tool for ophthalmologists to improve

diagnostic accuracy and efficiency in detecting glaucoma.

● Retinal specialists: Retinal specialists can utilize the system for a more objective

assessment of OCT scans in patients suspected of glaucoma.

● Glaucoma researchers: Researchers can leverage the system to develop and evaluate

new machine learning algorithms for glaucoma detection.

Dept. of ISE, MITE, Moodabidri Page 8

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

3.1.4 Design and Implementation Constraints

• Data Availability and Quality: The performance of the machine learning system heavily

relies on the quality and quantity of training data. Access to a large dataset of labeled OCT

images is crucial for effective training.

• Computational Resources: Training and running machine learning models, particularly

deep learning models, require significant computational power. Access to powerful computing

resources like GPUs is essential for efficient system development.

• Interpretability of Results: While machine learning models can be highly accurate,

ensuring the interpretability of their results is important. Understanding how the model arrives

at its predictions can be crucial for doctors in clinical decision-making.

• Regulatory Considerations: Regulatory approval processes for medical devices

incorporating machine learning algorithms need to be navigated for clinical use.

3.1.5 Assumptions and Dependencies

• The system will be integrated into existing medical imaging software used by

ophthalmologists.

• Ophthalmologists using the system will have a basic understanding of machine learning

and its limitations in medical diagnosis.

• The system will have access to a secure and reliable storage system for medical images

and patient data.

3.2 Specific Requirements

This section includes the detailed description about the hardware requirements, software
requirements, functional requirements and non-functional requirements.

3.2.1 Hardware Requirements

Hardware requirements refer to the physical parts of a computer and related devices. Internal
hardware devices include motherboards, hard drives and RAM External hardware devices
include monitors, keyboards, mice, printers, and scanner.

Dept. of ISE, MITE, Moodabidri Page 9

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

• Processor :1 GHZ or higher CPU

• Hard disk : 500 MB available internal storage
• Memory : 8 GB of RAM is minimally recommended
• Display : 2.8 inches or larger

3.2.2 Software Requirements

Software requirement is a field within Software Engineering that deals with establishing the
Deeds of stakeholders that are to be solved by the software. The software requirements of our
project given below
• Operating System : Windows 7 or above/ Linux
• Web Server : Flask
• Programming Languages : Python, CSS, JS, OpenCV
• Web browsers : Google chrome/Mozilla Firefox/internet explorer
• IDE : Visual studio code

1) Python

Python is a high-level, interpreted, general-purpose programming language. Its design

philosophy emphasizes code readability with the use of significant indentation. Python uses
whitespace indentation, rather than curly brackets or keywords, to delimit blocks. An increase
in indentation comes after certain statements; a decrease in indentation signifies the end of the
current block. Thus, the program's visual structure accurately represents its semantic structure.
This feature is sometimes termed the off-side rule. Some other languages use indentation this
way; but in most, indentation has no semantic meaning. The recommended indent size is four
spaces.

2) Flask

Flask is a micro web framework written in Python, designed to make web development quick
and easy while providing the flexibility to scale up to complex applications. It follows a
minimalistic approach, allowing developers to get started quickly without imposing strict
conventions. At the core of Flask are routes, which map URL patterns to Python functions,
known as views or endpoints. These views handle HTTP requests and generate responses,
allowing developers to define what happens when a user visits a specific URL. Flask uses Jinja2
templating engine to render HTML templates, separating the presentation logic from the

Dept. of ISE, MITE, Moodabidri Page 10

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

application logic. Request and response objects provide access to request data and allow
customization of HTTP responses. Flask applications can be configured using configuration
variables to control various aspects such as debug mode and database settings. Error handling
mechanisms handle exceptions and errors during request processing, and Flask's rich ecosystem
of extensions adds additional functionality like database integration, authentication, and form
validation. With its simplicity, flexibility, and extensibility, Flask remains a popular choice for
web development in Python.

3) Visual Studio Code

Visual Studio includes a code editor supporting IntelliSense as well as code refactoring. The
integrated debugger works both as a source-level debugger and a machine-level debugger.
Other built-in tools include a code profiler, designer for building GUI applications, web
designer, class designer, and database schema designer. It accepts plug-ins that expand the
functionality at almost every level—including adding support for source control systems (like
Subversion and Git) and adding new toolsets like editors and visual designers for domain-
specific languages or toolsets for other aspects of the software development lifecycle (like the
Azure DevOps client: Team Explorer). The most basic edition of Visual Studio, the Community
edition, is available free of charge. The slogan for Visual Studio Community edition is "Free,
fully-featured IDE for students, open-source and individual developers". As of March 2021, the
current production-ready Visual Studio version was 2019, with older versions such as 2012 and
2013 on Extended Support, and 2015 and 2017 on Mainstream Support On April 12, 2010,
Microsoft released Visual Studio 2010, codenamed Dev10, and .NET Framework 4.

4) HTML

Hypertext Markup Language (HTML) is the standard markup language for creating web pages
and web applications. With Cascading Style Sheets (CSS) and JavaScript, it forms a triad of
cornerstone technologies for the World Wide Web.

Web browsers receive HTML documents from a web server or from local storage and render
the documents into multimedia web pages. HTML describes the structure of a web page
semantically and originally included cues for the appearance of the document.HTML elements
are the building blocks of HTML pages. With HTML constructs, images and other objects such
as interactive forms may be embedded into the rendered page. HTML provides a means to
create structured documents by denoting structural semantics for text such as headings,
paragraphs, lists, links, quotes and other items. HTML elements are delineated by tags, written

Dept. of ISE, MITE, Moodabidri Page 11

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

using angle brackets. Tags such as <img /> and <input /> directly introduce content into the
page. Other tags such as <p> surround and provide information about document text and may
include other tags as sub-elements. Browsers do not display the HTML tags, but use them to
interpret the content of the page. HTML can embed programs written in a scripting language
such as JavaScript, which affects the behavior and content of web pages. Inclusion of CSS
defines the look and layout of content. The World Wide Web

Consortium (W3C), maintainer of both the HTML and the CSS standards, has encouraged the
use of CSS over explicit presentational HTML since 1997.

5) JavaScript

JavaScript often abbreviated as JS, is a high-level, interpreted programming language that

conforms to the ECMAScript specification. It is a programming language that is characterized
as dynamic, weakly typed, prototype-based and multi-paradigm. Alongside HTML and CSS,
JavaScript is one of the core technologies of the World Wide Web. JavaScript enables
interactive web pages and is an essential part of web applications. The vast majority of websites
use it, and major web browsers have a dedicated JavaScript engine to execute it. As a multi-
paradigm language, JavaScript supports event-driven, functional, and imperative (including
object-oriented and prototype-based) programming styles. It has APIs for working with text,
arrays, dates, regular expressions, and the DOM, but the language itself does not include any
I/O, such as networking, storage, or graphics facilities. It relies upon the host environment in
which it is embedded to provide these features.

6) Operating System: Microsoft Windows XP/Windows 7

Windows XP/7analyses the performance impact of visual effects and uses this to determine
whether to enable them, so as to prevent the new functionality from consuming excessive
additional processing overhead. Users can further customize these settings. Some newer video
cards. However, if the video card in not capable of hardware alpha blending, performance can
be substantially degraded, and Microsoft recommends the feature should be turned off
manually. Windows XP/7 added the ability for windows to use “Visual Styles” to change the
appearance of the user interface.

3.2.3 Functional Requirements

The system shall consist of the following modules:

Dept. of ISE, MITE, Moodabidri Page 12

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

• Data Acquisition Module: This module facilitates the import of OCT images from

various sources, ensuring compatibility with different imaging devices and file formats.

• Preprocessing Module: This module performs essential preprocessing steps on OCT

images to ensure consistency and improve analysis accuracy. Preprocessing techniques might

include noise reduction, normalization, and image registration.

• Feature Extraction Module: This module extracts relevant features from the

preprocessed OCT images. These features could include retinal nerve fiber layer thickness

measurements, cup-to-disk ratio, and other parameters indicative of glaucomatous changes.

• Machine Learning Module: This core module houses the machine learning model

responsible for analyzing extracted features and classifying OCT images. The system can be

designed to support various machine learning algorithms depending on project goals and data

availability.

• Classification Module: Based on the analysis of extracted features by the machine

learning model, this module classifies OCT images as either positive (glaucoma present) or

negative (glaucoma absent).

• Result Reporting Module: This module generates reports summarizing the system's

analysis of OCT images. These reports can be tailored to provide ophthalmologists with

relevant information for clinical decision-making.

3.2.4 Non-Functional Requirements

• Accuracy: The system should achieve a high degree of accuracy in classifying OCT

images for glaucoma detection. This translates to minimizing both false positives (healthy eyes

identified as glaucomatous) and false negatives (glaucomatous eyes missed by the system).

• Performance: The system should exhibit efficient processing times for analyzing OCT

images. This ensures minimal delays in delivering results to ophthalmologists.

• Scalability: The system should be designed to accommodate potential growth in data

volume and user base. This might involve implementing modular architecture and leveraging

cloud-based computing resources.

• Security: The system must adhere to strict security protocols to safeguard sensitive

patient data, including OCT images and medical records.

Dept. of ISE, MITE, Moodabidri Page 13

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

• Usability: The user interface should be intuitive and user-friendly for ophthalmologists

with varying levels of technical expertise.

Dept. of ISE, MITE, Moodabidri Page 14

 Chapter – 4
GANTT CHART
 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

CHAPTER 4

GANTT CHART

A Gantt chart is a type of bar chart, developed by Henry Gantt that illustrates a project schedule.
Gantt charts illustrate the start and finish of the terminal elements and summary elements of the
project. Terminal elements and summary elements comprise the work breakdown structure of
the project.

The following is the Gantt chart of the project “LIGHTNING PREDICTION AND ALERT
SYSTEM”

Figure 4.1 Gantt chart

Dept. of ISE, MITE, Moodabidri Page 15

 Chapter – 5
SYSTEM DESIGN
 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

CHAPTER 5

SYSTEM DESIGN
5.1 Architectural Diagram

An architecture diagram is a graphical representation of a set of concepts, that are part of

architecture, including their principles, elements and components. Sometimes, it’s helpful to
get a big picture view of something for context and an architectural diagram. An architecture
description is a formal description of a system, organized in a way that supports reasoning about
the structural properties of the system. It defines the system components or building blocks and
provides a plan from which products can be procured, and system developed, that will work
together to implement the overall system. Architectural diagram helps in:
• They help with comprehension: architectural diagrams help convey complex information
in a single image like there is a saying a picture is worth a thousand words.
• They improve communication and collaboration: when we are working on anything that
involves multiple people, there is always a risk of miscommunication between them, so it is
important to standardize information, which is where the architectural diagrams are very useful.

Fig 5.1 Architecture of the system

Dept. of ISE, MITE, Moodabidri Page 16

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

5.2 Use Case Diagram and Description

A use case is a set of scenarios that describing an interaction between a source and a
destination. The two main components of a use case diagram are use cases and actors. It
displays the relationship among them. A use case diagram can identify the different types of
users of a system and the different use cases and will often be accompanied by other types of
diagrams as well. The use cases are represented by either circles or ellipses. Use case diagram
of this system is as shown below.

Fig 5.2 Use Case design of the System

5.3 Sequence Diagram

A Sequence diagram shows how a set of objects communicate with each other to perform a
Complex task. This type of diagram allows the other developer to verify that the interaction
is correct. A Sequence diagram shows parallel vertical lines, different processes or objects
that live simultaneously and as horizontal arrows, the messages exchanged between them,
in the
order in which they occur. This allows the specification of simple runtime scenarios in a
Graphical manner.

Dept. of ISE, MITE, Moodabidri Page 17

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

The Fig. 5.3 represents the entire workflow of the proposed system. It shows the sequence in
which the system actually works.

Fig 5.3 Sequence Diagram of the System

5.4 Activity Diagram

Activity diagram is an important diagram in UML to describe the dynamic aspects of the
system. Activity diagram is basically a flowchart to represent the flow from one activity to
another activity. The activity can be described as an operation of the system.
The components used in this are as follows: Rounded Rectangle: It indicates the process. Arrow: It
indicates a transition line.
Rhombus: It indicates the decision.
Solid Circle: It represents the initial state of workflow.

Dept. of ISE, MITE, Moodabidri Page 18

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

Encircled Black Circle: It represents the final state of workflow.

Fig 5.4 Activity Diagram of the System

5.5 Data Flow Diagram

A Data Flow Diagram (DFD) is a graphical illustration of the flow of data throughout an
information system, modelling its procedural aspects. A DFD is often used as a beginning step
to create an outline of the system, which can later be elaborated. DFDs can also be used for the
visualization of data processing.

Dept. of ISE, MITE, Moodabidri Page 19

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

Fig 5.5 Data Flow Diagram of the System

Dept. of ISE, MITE, Moodabidri Page 20

 Chapter – 6
IMPLEMETATION
 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

CHAPTER 6

IMPLEMENTATION

6.1 System Implementation

Introduction
A crucial phase in the system development life cycle is successful implementation of new
system design. Implementations simply mean converting new system design into operation.
The term implementation has different meanings, ranging from the conversion of a basic
application to a complete replacement of computer system. Implementation is used here to
mean the process of converting a new or revised system design into an operational one.

6.1.1 Algorithm Used

The algorithms used in this project:
• Convolution Neural Network (CNN)

Convolution Neural Network

Convolutional neural network is a representative algorithm in deep learning. It is essentially

a multi-layer perceptron that simulates local perception to achieve an input-to-output
mapping. It extracts the characteristics of the data at different scales through multiple
convolutions and pooling. What is unique in the CNN network is the way used in local
connections and shared weights. On the one hand, it reduces the number of weights which
makes the network easy to optimize, and on the other hand, it reduces the risk of overfitting.
CNNs are generally composed of three mutually supported levels, namely convolutional layer,
pooling layer, fully connected and Softmax layer. In the convolution process, we get local
features. Since one of the convolution layers is composed of multiple convolution units, in the
calculation process, in order
Dept. of ISE, MITE, Moodabidri Page 21
 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

to extract more features about the input parameters, it is necessary to obtain more complex
feature correlation values from low-level convolutional layers through multi-level cascading.

6.2 Workflow

The following steps are carried out to implement the proposed system.

6.2.1 Pre-processing
Data Pre-processing is a technique that is used to convert the raw data into a clean data set. In
other words, whenever the data is gathered from different sources it is collected in raw format
which is not feasible for the analysis.
Dataset is pre-processed to resize the dataset images. We created our own dataset from a
lightning video and created frames and images are resized to 150*150 before processing them
further.

6.2.2 Train Test Split

The process of organizing data into groups and classes on the basis of certain characteristics
is known as the classification of data. Classification helps in making comparisons among the
categories of observations. It can be either according to numerical characteristics or according
to attributes. So here we need to visualize the prepared data to find whether the training data
contains the correct label, which is known as a target or target attribute. In this project the
dataset is split into 90 samples for training and 10 for testing.

6.2.3 Training
The process of training an ML model involves providing an ML algorithm (that is, the learning
algorithm) with training data to learn from. The term ML model refers to the model artifact
that is created by the training process. The training data must contain the correct answer, which
is known as a target or target attribute. The learning algorithm finds patterns in the training
data that map the input data attributes to the target (the answer that you want to predict), and
it outputs an ML model that captures these patterns.

6.2.4 Picking the Model

Pickling is a process in which model is stored in a file for future use. Before pickling model
must be trained well and optimized to maximum extent possible because after picking model,

Dept. of ISE, MITE, Moodabidri Page 22

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT
data will not be trained. By pickling there is no need of training the model each time the user
makes a classification.

6.2.5 Pseudo code of system

Step 1: Collect dataset

- Obtain a dataset of OCT images containing both normal and glaucomatous cases.

Step 2: Pre-process dataset

- Import necessary libraries/modules (e.g., OpenCV, numpy)
- For each image in the dataset:
- Read the image using imread()
- Pre-process the image (e.g., resize, normalization)
- Perform any additional pre-processing steps if necessary

Step 3: Train YOLOv8 Model

- Download or implement YOLOv8 model for object detection
- Prepare data for YOLOv8 training (annotate images, create train/validation splits)
- Train the YOLOv8 model on the annotated dataset

Step 4: Extract Features

- Use the trained YOLOv8 model to detect regions of interest (ROIs) in OCT images
- Extract features from the detected ROIs (e.g., texture features, intensity features)

Step 5: Build Machine Learning Model

- Choose a suitable machine learning algorithm (e.g., SVM, Random Forest)
- Prepare the feature vectors and corresponding labels
- Split the dataset into training and testing sets

Step 6: Train Machine Learning Model

- Train the machine learning model using the training dataset
- Tune hyperparameters if necessary to optimize performance

Step 7: Evaluate Model

- Evaluate the trained model using the testing dataset
- Calculate metrics such as accuracy, precision, recall, and F1-score

Step 8: Fine-tune Model (Optional)

- If model performance is not satisfactory, consider fine-tuning the YOLOv8 model or
adjusting machine learning parameters

Step 9: Save Model

- Save the trained machine learning model and YOLOv8 weights for future use

Step 10: Glaucoma Prediction

- Use the trained model to predict glaucoma in new OCT images
- Display or output the prediction results

6.2.6 Code implemented

The following are the code snippets used in the project.

Dept. of ISE, MITE, Moodabidri Page 23

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

• Importing required packages

• Exporting Dataset

• Custom Training

Dept. of ISE, MITE, Moodabidri Page 24

Chapter – 7
TESTING
 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

CHAPTER 7

TESTING

7.1 INTRODUCTION

Testing is a crucial stage in the development of a machine learning system for glaucoma

detection using OCT images. It ensures the system functions as intended, delivers accurate

results, and meets quality standards. There are various testing methodologies employed to

achieve comprehensive evaluation.

7.2 Testing Approaches

There are two main testing approaches used in software development:

• White-Box Testing (Glass Box Testing): This approach delves into the system's internal
workings, code structure, and logic. Testers with knowledge of the underlying algorithms and
implementation details design test cases that target specific code segments, branches, and
internal functionalities.
o Advantages:
▪ Efficient test case design based on code knowledge.
▪ Promotes code optimization and removal of redundant code.
o Disadvantages:
▪ Requires skilled testers with expertise in the system's code.
▪ Limited ability to uncover unforeseen errors not explicitly targeted by test cases.
• Black-Box Testing (Functional Testing): This approach focuses on the system's external
behavior and functionality without delving into its internal structure. Testers design test cases
based on the system's requirements and specifications. These cases involve providing valid
and invalid inputs to evaluate the system's expected outputs and responses.
o Advantages:
▪ Reproducible test cases for consistent evaluation.
▪ Tests the system's behavior in its intended operating environment.
▪ Cost-effective due to less reliance on specialized code knowledge.
o Disadvantages:
▪ Overestimation of testing coverage due to the inability to test all internal functionalities.
▪ Difficulty in pinpointing the root cause of failures identified during testing.

Dept. of ISE, MITE, Moodabidri Page 25

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT
7.3 Test Levels

Testing is conducted at various levels throughout the development process:

● Unit Testing: Individual components of the machine learning system, such as modules

responsible for image preprocessing or feature extraction, are tested in isolation to ensure they

function as designed.

● Integration Testing: Once individual components are validated, they are integrated into

a cohesive system. Integration testing verifies if these components interact seamlessly and

produce the expected overall functionality.

● System Testing: The complete system, including the user interface and interaction with

OCT image data, is rigorously evaluated to ensure it adheres to overall system requirements

and delivers accurate results for glaucoma detection.

7.4 Evaluation Metrics

Beyond functional testing, the system's performance is assessed using specific metrics:

• Accuracy: This metric measures the system's ability to correctly classify OCT images,

differentiating between healthy and glaucoma-affected eyes. It is typically expressed as a

percentage of correctly classified images.

• Sensitivity: This metric indicates the system's ability to correctly identify glaucoma cases

(positive predictive value). A high sensitivity minimizes false negatives (missed glaucoma

cases).

• Specificity: This metric reflects the system's ability to correctly identify healthy eyes

(negative predictive value). A high specificity minimizes false positives (healthy eyes

identified as glaucoma).

• Precision: This metric represents the proportion of positive test results that are truly

positive (true positives divided by the sum of true positives and false positives).

• Recall: This metric represents the proportion of positive cases that are correctly identified

(true positives divided by the sum of true positives and false negatives).

Dept. of ISE, MITE, Moodabidri Page 26

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT
7.5 Test Cases: Prediction Functionality

These test cases focus on evaluating the core functionality of the system: uploading an OCT

image and receiving a glaucoma detection prediction.

Test Case 1: Uploading a Valid OCT Image

● Action: User uploads a valid OCT image file (e.g., JPG, PNG)

● Expected Output:

○ System successfully uploads the image.

○ System processes the image and displays the glaucoma detection prediction (e.g.,

"Glaucoma detected" or "Glaucoma not detected").

● Actual Output: Classified results are obtained.

Test Case 2: Uploading an Unsupported File Type

● Action: User uploads a file with an unsupported format (e.g., DOCX, PDF)

● Expected Output:

○ System displays an error message indicating the file format is not supported.

○ The upload process is cancelled.

● Actual Output: Result not displayed (Method not allowed Error).

Test Case 3: Uploading a Corrupted File

● Action: User uploads a corrupted or damaged image file

● Expected Output:

○ System displays an error message indicating the file is corrupted or invalid.

○ The upload process is cancelled.

● Actual Output: Result not displayed (Method not allowed Error).

Test Case 4: Empty Upload Attempt

● Action: User clicks the upload button without selecting an image file

● Expected Output:

○ System displays an error message prompting the user to select an image file.

Dept. of ISE, MITE, Moodabidri Page 27

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT
○ The upload process is not initiated.

● Actual Output: Undefined classification.

Dept. of ISE, MITE, Moodabidri Page 28

 Chapter – 8
RESULT AND SNAPSHOTS
 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

CHAPTER 8

RESULTS AND SNAPSHOTS

Fig 8.1 Home page

Fig 8.2 About Us page

Dept. of ISE, MITE, Moodabidri Page 29

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

Fig 8.3 Blog page

Fig 8.4 Classifier page

Dept. of ISE, MITE, Moodabidri Page 30

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

Fig 8.5 Contact Page

Dept. of ISE, MITE, Moodabidri Page 31

 Chapter – 9
CONCLUSION AND FUTURE
WORK
 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

CHAPTER 9

CONCLUSION

In conclusion, this project pushes the boundaries of glaucoma detection by integrating

YOLOv8, a powerful object detection model from the leading edge of deep learning. This
innovative approach departs from traditional methods that rely on feature extraction, allowing
the system to directly analyze OCT images for subtle signs of glaucoma. This streamlining not
only enhances efficiency but also prioritizes exceptional accuracy in classifying healthy and
glaucomatous eyes. The emphasis on user-friendliness throughout the design translates to a
system that seamlessly integrates into existing clinical workflows. Ophthalmologists can
leverage functionalities like data import from various imaging devices, meticulous data
preprocessing to ensure optimal analysis, and clear result reporting to aid in informed clinical
decision-making. Furthermore, the project prioritizes reliability and security. By adhering to
strict non-functional requirements, the system guarantees a high degree of accuracy in OCT
image classification while minimizing analysis time. Robust security protocols safeguard
sensitive patient data, fostering trust and ethical implementation in clinical settings. This
systematic approach, encompassing meticulous data collection, preprocessing, and model
development using YOLOv8, paves the way for a dependable and user-friendly system for
glaucoma detection. This advancement holds immense potential to transform ophthalmic
diagnostics. By enabling earlier and more accurate diagnoses, the system can significantly
improve treatment planning and patient outcomes, ultimately contributing to better vision health
for all.

Dept. of ISE, MITE, Moodabidri Page 32

 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

FUTURE WORK

This project lays a strong foundation for further advancements in glaucoma detection using

machine learning. Here are some key areas for future exploration:

● Real-time Analysis: Transitioning the system from an offline analysis tool to a real-time
system would significantly enhance its clinical utility. This would involve optimizing
algorithms and system architecture for efficient processing of OCT images during patient
consultations.

● Integration with Diagnostic Devices: Integrating the system with ophthalmic imaging
devices could streamline the diagnostic workflow. This would allow for seamless transfer of
OCT images directly from the device for immediate analysis by the system, potentially reducing
analysis time and improving overall efficiency.

● Incorporation of Additional Data: Exploring the inclusion of additional data sources,

such as patient demographics, medical history, and other ophthalmic test results, could
potentially improve the system's diagnostic accuracy. By leveraging a more comprehensive data
profile, the system could learn complex relationships between various factors and glaucoma
risk.

● Explainable AI Integration: Integrating explainable AI (XAI) techniques could enhance

the interpretability of the system's predictions. This would allow ophthalmologists to
understand the reasoning behind the system's diagnosis, fostering trust and transparency in
clinical decision-making.

● Multimodal Learning: Exploring multimodal learning approaches that combine OCT

images with other modalities, such as visual field tests, could provide a more holistic view of
glaucoma progression. This could potentially improve the system's ability to detect early-stage
glaucoma and other related eye conditions.

● Continuous Model Improvement: Continuously monitoring and refining the machine

learning model through additional data collection and model retraining can ensure the system
maintains exceptional accuracy and adapts to evolving clinical needs.

Dept. of ISE, MITE, Moodabidri Page 33

REFERENCES
 GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

REFERENCES
1) https://tvst.arvojournals.org/solr/searchresults.aspx?author=Jie+Wang
2) https://iovs.arvojournals.org/solr/searchresults.aspx?author=Filippo
3) https://www.mdpi.com/2079-9292/9/10/1617
4) https://pubmed.ncbi.nlm.nih.gov/37565126/
5) https://doi.org/10.3390/s22062342
6) https://doi.org/10.1016/j.jksuci.2021.12.002
7) https://doi.org/10.3390/healthcare11020212
8) https://rdcu.be/dCUsd

Dept. of ISE, MITE, Moodabidri Page 34

 ISSN (Online) 2278-1021

IJARCCE ISSN (Print) 2319-5940

GLAUCOMA DETECTION USING MACHINE LEARNING WITH OCT

International Journal of Advanced Research in

Computer and Communication Engineering

A monthly Peer-reviewed journal

Indexed by Google Scholar, Crossref and Mendeley

DOI IJARCCE/10.17148

Dept. of ISE, MITE, Moodabidri Page 35

www.ijarcce.com
 IJARCCE ISSN (O) 2278-1021, ISSN (P) 2319-5940

International Journal of Advanced Research in Computer and Communication Engineering

Impact Factor 8.102Peer-reviewed & Refereed journalVol. 13, Issue 3, March 2024
DOI: 10.17148/IJARCCE.2024.133117

Glaucoma Detection using Machine Learning

with OCT
Sharath Kumar1, Athik Rehaman2, Lathesh Kumar3, Mayur S Karkera4,
Mohammed Muneef5
Assistant Professor, Dept. of Information Science & Engineering, Mangalore Institute of Technology & Engineering,
Moodabidre, India1
Student, Dept. of Information Science & Engineering, Mangalore Institute of Technology & Engineering,
Moodabidre, India2-5

Abstract: Glaucom Detection is devoted to advancing the diagnosis of glaucoma and its primary sub-conditions:
Choroidal Neovascularization (CNV), Diabetic Macular Edema (DME), and Drusen, recognizing their significant impact
on global vision loss. Glaucoma, often termed the "silent thief of sight," leads to progressive optic nerve damage and
irreversible vision loss. While elevated intraocular pressure (IOP) is a primary cause, not all cases stem from high IOP,
underscoring the multifactorial nature of its development involving genetics, age, race, and family history. This
complexity necessitates a comprehensive approach for early detection and management to counteract the disease's silent
progression.To meet the pressing need for early intervention, this project integrates cutting-edge machine learning and
image analysis techniques alongside traditional diagnostic methods like tonometry and ophthalmoscopy. By harnessing
these advanced technologies, the project aims to enhance early detection capabilities, facilitating tailored approaches to
diagnose and manage glaucoma and its sub-conditions effectively. Early identification is paramount due to the insidious
nature of glaucoma, which often advances unnoticed until irreversible damage occurs. Prioritizing early intervention not
only decelerates disease progression but also safeguards the quality of life for affected individuals, underscoring the
significance of personalized diagnostic and treatment strategies for various complications and subtypes of glaucoma.
This holistic approach seeks to revolutionize glaucoma management by integrating state-of-the-art technology with
established diagnostic methods, ultimately improving outcomes for patients worldwide..

Keywords: Glaucoma diagnosis Choroidal Neovascularization (CNV) Diabetic Macular Edema (DME) Drusen Early
detection Machine learning Image analysis.

I. INTRODUCTION

Glaucoma, often referred to as the "silent thief of sight," is a prevalent and potentially devastating eye disease
characterized by the progressive damage to the optic nerve, leading to irreversible vision loss. It is estimated that over 80
million individuals worldwide are affected by this condition, making it one of the leading causes of blindness. While
glaucoma primarily affects the elderly population, it can also manifest in individuals of all ages, making its early detection
and management of paramount importance. This project is dedicated to the development and enhancement of diagnostic
methods for glaucoma, with a specific focus on the identification of its primary sub-conditions: Choroidal
Neovascularization (CNV), Diabetic Macular Edema (DME), and Drusen.

To fully comprehend the significance of these sub-conditions and the urgency for their early detection, we must delve
into the causes and symptoms of glaucoma itself.he primary cause of glaucoma is elevated intraocular pressure (IOP), a
condition where the fluid within the eye, known as aqueous humor, is not adequately drained. This leads to increased
pressure within the eye, which, over time, damages the optic nerve.

Elevated IOP is a major risk factor, but it is essential to recognize that not all individuals with glaucoma have high IOP,
and not all individuals with high IOP develop glaucoma. Additionally, other factors such as genetics, family history, age,
and race can play a role in glaucoma's development.While traditional diagnostic methods such as tonometry and
ophthalmoscopy are valuable, this project aims to employ state-of-the-art machine learning and image analysis techniques
to enhance the early detection of glaucoma and its sub conditions, CNV, DME, and Drusen. By advancing our
understanding and diagnostic capabilities, we endeavor to contribute to improved patient care, early intervention, and the
preservation of vision The project's emphasis on early detection is crucial because glaucoma and its sub-conditions often
progress silently, causing irreversible damage before noticeable symptoms manifest.

© IJARCCE This work is licensed under a Creative Commons Attribution 4.0 International License 729
 IJARCCE ISSN (O) 2278-1021, ISSN (P) 2319-5940

International Journal of Advanced Research in Computer and Communication Engineering

Impact Factor 8.102Peer-reviewed & Refereed journalVol. 13, Issue 3, March 2024
DOI: 10.17148/IJARCCE.2024.133117
Early intervention and treatment can significantly slow the progression of the disease, preserving the quality of life for
affected individuals. Moreover, the inclusion of CNV, DME, and Drusen in this project underscores the need to address
specific complications and subtypes of glaucoma, tailoring diagnostic and treatment approaches to individual patient
needs.

II. LITERATURE SURVEY

In [1] Jie Wang, et.al., Researchers developed a deep learning algorithm for diagnosing and segmenting choroidal
neovascularization (CNV) in OCTA images. This algorithm showed high accuracy, potentially improving CNV diagnosis
and treatment. It can automate CNV analysis, reducing workload for ophthalmologists and improving patient care.

In [2] Filippo Arcadu, et.al., This study used deep learning to analyze retinal photographs (CFPs) to predict diabetic
macular thickening (MT) normally measured with retinal scans (OCT). The models achieved high accuracy and could
improve diagnosis efficiency in telemedicine. However, real-world validation is needed.

In [3] Quang T.M.Pham, et.al., This paper proposes a new deep learning model for segmenting drusen, deposits linked
to age-related macular degeneration (AMD). Unlike prior methods using single-scale analysis, this model incorporates
both global and local image information.This research shows promise for applying deep learning in AMD diagnosis.

In [4] Rakesh Patel, et.al., This study shows a deep learning model can effectively detect drusen, CNV, and DME in OCT
images. This offers a more standardized approach to diagnosing AMD and DR. More research is needed before
widespread use, but this technology has promise to improve patient care, especially in areas with limited access to
specialists.

In [5] Ibrahim Yasser, et.al., This is an article about automated diagnosis of Optical Coherence Tomography Angiography
(OCTA). It discusses using machine learning techniques to diagnose retinal diseases. The article proposes a new method
to classify OCTA images based on convolutional neural networks (CNNs). The method involves dividing the OCTA
image into smaller pieces and cropping a specific region around the fovea.

In [6] Pawan Kumar Upadhyay, et.al.,Researchers have developed a new and powerful neural network for analyzing
retinal diseases using optical coherence tomography (OCT) images. This network can effectively classify four distinct
retinal conditions: neovascularization, diabetic macular edema, drusen, and normal.

In [7] Amit Choudhary, et.al.,This study utilizes a customized 19-layer VGG-19 deep learning model to automatically
detect retinal diseases from OCT images. Trained on a large public dataset, the model classifies four conditions: choroidal
neovascularization, drusen, diabetic macular edema, and normal.

In [8] Zhenhua Wang , et.al., This is an article about automated diagnosis of Optical Coherence Tomography
Angiography (OCTA). It discusses using machine learning techniques to diagnose retinal diseases. The article proposes
a new method to classify OCTA images based on convolutional neural networks (CNNs). The method involves dividing
the OCTA image into smaller pieces and cropping a specific region.

In [9] Nawaaz A Nathoo, et.al., This is an article about measuring drusen load to predict age-related macular degeneration
(AMD). It discusses a study that used spectral-domain optical coherence tomography (SD OCT) to measure drusen load
in patients with non-neovascular AMD. The study found that drusen load was associated with development of RPE
atrophy and neovascular AMD.

III. SCOPE AND METHODOLOGY

Aim of the project

This project tackles the challenge of early glaucoma detection and accurate subclassification. Glaucoma, often
progressing silently, requires prompt intervention for vision preservation. Current methods rely on subjective analysis of
retinal features, potentially limiting accuracy. Furthermore, distinguishing glaucoma subtypes like CNV, DME, and
Drusen can be difficult. Our solution lies in developing an automated system that analyzes OCT images using a YOLOv8
model under the CNN architecture. This model will be trained to identify specific features in the optic nerve and retinal
layers, allowing for not only glaucoma detection but also differentiation of its subtypes and classification of normal OCT
images. This project aspires to equip ophthalmologists with a powerful tool, potentially leading to a future with improved
glaucoma detection, accurate subclassification, and empowered patients who can proactively manage their eye health.

© IJARCCE This work is licensed under a Creative Commons Attribution 4.0 International License 730
 IJARCCE ISSN (O) 2278-1021, ISSN (P) 2319-5940

International Journal of Advanced Research in Computer and Communication Engineering

Impact Factor 8.102Peer-reviewed & Refereed journalVol. 13, Issue 3, March 2024
DOI: 10.17148/IJARCCE.2024.133117
Existing system
While previous efforts have explored automated glaucoma detection using OCT images, achieving accurate classification
of glaucoma subtypes remains a challenge. This project builds upon these advancements by incorporating a transfer
learning approach. We'll utilize a pre-trained Convolutional Neural Network (CNN) called VGG-19. This powerful
model, trained on a vast dataset of general images, possesses valuable image recognition skills. By leveraging VGG-19's
pre-trained knowledge as a foundation, we aim to develop a more efficient system. During "fine-tuning," we'll adjust
VGG-19's weights to focus on recognizing specific features within OCT scans that differentiate drusen, CNV, DME, and
normal retinas.It is also important to acknowledge the limitations associated with transfer learning. The pre-trained model
might have inherent biases based on the original dataset (ImageNet), potentially requiring extensive fine-tuning with a
large and diverse collection of OCT images to achieve optimal performance for glaucoma subclassification.

Proposed system
Building upon advancements in automated glaucoma detection with OCT images, this project seeks a more accurate
solution for classifying glaucoma subtypes. While existing systems utilize transfer learning with models like VGG-19,
these approaches can be limited by biases inherent in the pre-trained datasets used for general image recognition. To
address this, we propose a novel approach using YOLOv8, a state-of-the-art object detection model built on the CNN
architecture. Unlike VGG-19, YOLOv8 is specifically designed for object detection tasks, making it a more natural fit
for identifying crucial features within OCT scans, such as the optic nerve and retinal layers. Additionally, by training
YOLOv8 from scratch on a specifically curated dataset of OCT images labeled for glaucoma, CNV, DME, Drusen, and
normal retinas, we aim to minimize potential biases and achieve superior accuracy in classifying these glaucoma
subtypes. This targeted approach using YOLOv8 has the potential to significantly improve early detection and
subclassification compared to previous methods that rely on transfer learning.

Fig 1. Proposed system

© IJARCCE This work is licensed under a Creative Commons Attribution 4.0 International License 731
 IJARCCE ISSN (O) 2278-1021, ISSN (P) 2319-5940

International Journal of Advanced Research in Computer and Communication Engineering

Impact Factor 8.102Peer-reviewed & Refereed journalVol. 13, Issue 3, March 2024
DOI: 10.17148/IJARCCE.2024.133117
System Architecture

A formal explanation of a system is provided by an architectural explanation, which is arranged to support reasoning
regarding the system's structure, individual components' properties that can be observed from the outside, and the
interactions between them. It also offers a framework from which systems can be developed and products acquired that
will cooperate to implement the system as a whole.

Fig 2. System Architecture

IV. CONCLUSION

In conclusion, this project ventured beyond the limitations of transfer learning approaches to explore the potential of
YOLOv8, a state-of-the-art object detection model built upon the robust CNN architecture. This shift towards a model
specifically designed for object detection tasks addresses the inherent biases that can plague pre-trained models used in
transfer learning. By leveraging YOLOv8's object detection prowess, we aim to develop a system that can meticulously
identify and differentiate the subtle features within OCT scans that distinguish between glaucoma, CNV, DME, Drusen,
and normal retinas. This targeted approach, fueled by a meticulously curated training dataset of labeled OCT images, has
the potential to significantly improve early detection and accurate classification of glaucoma subtypes compared to
existing methods. The potential benefits of this project extend far beyond improved classification accuracy. Earlier
diagnosis of glaucoma and its subtypes, enabled by this system, can lead to timely intervention and potentially vision
preservation for patients. Additionally, YOLOv8's ability to provide a more consistent and objective assessment
compared to traditional methods holds promise for enhanced diagnostic accuracy. Furthermore, streamlined workflows
for ophthalmologists achieved through automated analysis can translate into more time dedicated to patient care and
treatment planning. Ultimately, this project aspires to empower patients by enabling earlier detection, which allows them
to take a proactive stance in managing their eye health. By harnessing the power of deep learning and artificial
intelligence, this project lays the groundwork for a transformative future in glaucoma detection. A future where the
process becomes more efficient, accurate, and ultimately leads to better patient outcomes for those facing the challenges
of glaucoma. This project represents a significant step towards a brighter future for countless individuals whose vision is
threatened by this disease.

REFERENCES

[1] https://tvst.arvojournals.org/solr/searchresults.aspx?author=Jie+Wang
[2] https://iovs.arvojournals.org/solr/searchresults.aspx?author=Filippo+ Arcadu
[3] https://www.mdpi.com/2079-9292/9/10/1617
[4] https://pubmed.ncbi.nlm.nih.gov/37565126/
[5] https://doi.org/10.3390/s22062342
[6]https://doi.org/10.1016/j.jksuci.2021.12.002
[7]https://doi.org/10.3390/healthcare11020212
[8]https://rdcu.be/dCUsd
[9]10.1016/j.ajo.2014.06.021

© IJARCCE This work is licensed under a Creative Commons Attribution 4.0 International License 732