DETECTION OF DIABETIC RETINOPATHY

USING DEEP LEARNING

Submitted by: Haritha T H

LMC22MCA-2012
 INTRODUCTION
 Diabetic Retinopathy (DR) is a complication caused by diabetes that affects the human
eye. It is caused by the mutilation of the blood vessels of the light-sensitive tissue at the
back of the human retina.
 It’s the most recurrent cause of blindness in the working-age group of people and is
highly likely when diabetes is poorly controlled. Although, methods to detect Diabetic
Retinopathy exist, they involve manual examination of the retinal image by an
Ophthalmologist.
 Diabetic retinopathy is a leading cause of vision impairment and blindness among
individuals with diabetes. Early detection and intervention are crucial to prevent
irreversible damage to the retina.
 This project focuses on the development of an accurate and efficient system for the
detection of diabetic retinopathy using deep learning techniques.
 This project contributes to the ongoing efforts in leveraging artificial intelligence for the
improvement of eye care and the prevention of vision loss in diabetic patients.
 EXISTING SYSTEM
 The existing system likely relies on traditional methods, and transitioning to
deep learning could enhance accuracy and efficiency.
 A common or diabetic retinopathy detection, clinicians relied on traditional
ophthalmic examinations. These examinations involved dilating the patient's
pupils and using specialized instruments to directly view the retina. However,
this manual approach was labor intensive and subject to human error.
 The existing system, as outlined in the referenced IEEE paper, primarily
focuses on the development and evaluation of a deep learning model for
Diabetic Retinopathy (DR) detection. The model is trained using Resnet 18
and GoogleNet , achieving an accuracy of approximately 81%.
 PROPOSED SYSTEM
 Advancements in technology introduced fundus photography as a non-invasive and more
accessible method. It allowed for the documentation and analysis of retinal images,
paving the way for the development of computer-assisted systems to aid in diabetic
retinopathy diagnosis. This project focuses on the development of an accurate and
efficient system for the detection of diabetic retinopathy using deep learning techniques.
 The proposed methodology employs convolutional neural networks (CNNs) to
automatically extract meaningful features from digital fundus images. These images are
acquired through retinal screening, and the deep learning model is trained to classify
them into different stages of diabetic retinopathy, ranging from mild to severe.
 The project aims to compare the performance of Inception V3 and ResNet50 in terms of
accuracy, sensitivity, and specificity for DR detection and the most flexible one is
chosen. Here the fundus image is classified into several stages of DR such as normal,
mild, moderate, severe, and proliferative diabetic retinopathy.
 The project extends beyond model development, incorporating a user-
friendly web application with ReactJS for the frontend and a firebase
backend. Building upon the foundation of diabetic retinopathy (DR)
detection through deep learning, the proposed system is a comprehensive
solution that seamlessly integrates cutting-edge technology with user-
centric features.
 By leveraging convolutional neural networks (CNNs), specifically
comparing Inception V3 and ResNet50, the system aims to enhance the
accuracy, sensitivity, and specificity of DR detection. The project extends
beyond model development, incorporating a user-friendly web application
with ReactJS for the frontend and a firebase backend.
 A diverse and representative dataset, encompassing a variety of retinal
conditions and severity levels, is used for the training and validation of the deep
learning model. The model's performance is evaluated using standard metrics
such as sensitivity, specificity, accuracy, and area under the receiver operating
characteristic curve (AUC-ROC).
ALGORITHMS
 ResNet50
 ResNet-50 is a deep convolutional neural network architecture that was
introduced in 2015 as part of the ResNet (Residual Network) family of
models. ResNet-50 is specifically designed for image classification tasks and
has been widely used in various computer vision applications, including
object recognition, image segmentation, and medical image analysis.
 ResNet-50 is a 50-layer convolutional neural network (48 convolutional
layers, one MaxPool layer, and one average pool layer). Residual neural
networks are a type of artificial neural network (ANN) that forms networks
by stacking residual blocks.
 The 50-layer ResNet uses a bottleneck design for the building block. A
bottleneck residual block uses 1×1 convolutions, known as a “bottleneck”,
which reduces the number of parameters and matrix multiplications. This
enables much faster training of each layer. It uses a stack of three layers
rather than two layers.
It contains different layers, including convolutional,
pooling, and fully connected layers. This model can be
used as a feature extractor for the brain tumor
detection task.
The lower layers of the ResNet50 model learn public
image features that can be useful for brain tumor
detection. The last few layers of the ResNet50 model
are replaced by a new set of fully connected layers for
the specific task of brain tumor detection and
classification.
 Inception V3
 The Inception V3 is a deep learning model based on Convolutional Neural
Networks, which is used for image classification. The inception V3 is a
superior version of the basic model Inception V1 which was introduced as
GoogLeNet in 2014.
 The inception v3 model was released in the year 2015, it has a total of 42
layers and a lower error rate than its predecessors.
 Inception v3 is an image recognition model that has been shown to attain
greater than 78.1% accuracy on the ImageNet dataset. The model is the
culmination of many ideas developed by multiple researchers over the
years.
 The model itself is made up of symmetric and asymmetric building blocks,
including convolutions, average pooling, max pooling, concatenations, dropouts,
and fully connected layers. Batch normalization is used extensively throughout
the model and applied to activation inputs. Loss is computed using Softmax.
TOOLS USED
Operating System: Windows 11
Platform: Google collab
Language: Python 3.11
Framework: React
Backend: Firebase
IDE: VS Code
Libraries Used: Numpy, Flask, Pandas, Keras
PROCESS FLOW
DATABASE DESIGN
 LOGIN TABLE

FIELD TYPE

e-mail Varchar(225)

username Varchar(225)

password Varchar(225)

 HISTORY TABLE

FIELD TYPE

e-mail Varchar(225)

username Varchar(225)

password Varchar(225)
SCREENSHOTS
SCREENSHOTS
Thank you