Scribd
Upload
13 views22 pages

Final

Uploaded by

Sai tarun Maryala
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
13 views22 pages

Final

Uploaded by

Sai tarun Maryala
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 22

CNN-based Classification of

Diabetic Retinopathy Stages in


Fundus Images

TEAM - P7132
Team Members
Krithik Kethan Pujari
222010322009 GUID
E
Aryan Sahu
22201032201
3
Dr. Nandita
Bhanja
Chaudhuri
Ajay Esakki
2220103220
39
Sai Tarun Maryala
222010323009
Content
s
Aim Architecture
Objective
Implementatio
s
Abstract
Result n
Existing
Progres
Architecture
Literature s
Conclusion
Survey
Reference
AIM OF THE
PROJECT
The objective of this project is to leverage convolutional neural
network (CNN) architecture with TensorFlow/Keras to develop a
robust system for the early detection and precise classification
of diabetic retinopathy, advancing diagnostic capabilities in
healthcare.
Objective
s
Develop a convolutional neural network
(CNN) model using TensorFlow/Keras to
Implement data preprocessing techniques
such as rescaling, image augmentation, and
automatically detect diabetic filtering to enhance the quality and diversity
retinopathy in images of the training dataset.

Enhance the scalability of disease detection Finally, create a precise machine learning model
for classifying the severity of diabetic retinopathy
efforts by implementing technology-driven
on a scale from 0 to 4. This model will serve as a
solutions, aiming to alleviate the burden on valuable tool for clinicians, aiding in early
highly trained medical personnel. detection and intervention, and categorizing
severity levels from 'No DR' to 'Proliferative DR.'
Project Overview

Abstract
This project is employed to address the critical health
challenge posed by diabetic retinopathy (DR), aiming to

1 provide a reliable tool for classifying DR severity levels


and facilitating prompt clinical decision-making.Diabetic
retinopathy (DR) is a significant health challenge with
widespread prevalence and potential vision-threatening

Model
consequences. Timely detection is crucial for effective
intervention and halting the progression of DR.

Methodological Approach

Our project adopts a methodological approach using


Convolutional Neural Networks (CNNs) to automatically

2 extract features from retinal images, improving the


model's discernment of diabetic retinopathy severity
levels. The training process involves optimizing the CNN
architecture with a diverse dataset and rigorous
validation techniques for robust performance.

User-Friendly Interface and Benefi ts for


Diabetic Retinopathy Diagnosis:
"Our intuitive interface empowers clinicians to swiftly assess

3 diabetic retinopathy severity levels with automated scores,


expediting diagnostics, optimizing resources, and ensuring
fair, accurate evaluations for all patients."
LITERATURE
• The REVIEW
literature survey for the Diabetic Retinopathy project
involves reviewing existing research on the application of
Convolutional Neural Networks (CNNs) in the automated
detection and classifi cation of diabetic retinopathy from
fundus images.
• Previous studies emphasize the signifi cance of early
diagnosis using machine learning techniques and highlight
the challenges of classifying retinal images into diff erent
severity levels.
• Techniques such as image preprocessing, data augmentation
etc are used and Gaps in current research underscore the
need for robust, scalable, and interpretable models for
eff ective diabetic retinopathy screening.
Existing
System

U-Net ResNet Support Vector


Machine

Drawbacks: Drawbacks: Drawbacks:


• Poor feature • Limited Representation • Dependency on Feature
representation from the of Image Features Extraction.
initial layer. • Overfitting. • High Sensitivity to Noise
• Inability to Capture • Sensitivity to Small and Variability.
Long-Range Variations. • Overfitting.
Dependencies • Need for Manual Feature
• Sensitivity to Engineering.
Hyperparameters.
Our
Architecture
IMPLEMENTATION

Link to the Cola


b
Accuracy of the Model
RESULT

Representation of
Model Accuracy

Number of Epochs
Value of Loss function
RESULT

Representation of
Model Loss

Number of Epochs
Software Components
Python
The entire project is
Integrated
Development Librarie Dataset
implemented in
Python, so we need a
Environment
(IDE) s APTOS-2019
Blindness
Python interpreter Jupyter Book, TensorFlow,
installed on your
dataset
Colab OpenCV,
computer.
Pandas,
Hardware Components Matplotlib, tqdm

Computer Storage GPU Operating


Access to a powerful
system
Computer with suffi cient
processing power and Minimum 8 GB GPU as can
memory to run the disk space to signifi cantly speed up Windows 7 or newer,
model training and download and the training process 64-bit macOS 10.13+,
provide accurate results since the dataset is
install. or Linux
large
Visualization
Timeline
Project Planning and
Research
Data Collection and
Preprocessing
Model Development and
Training
Evaluation and
Validation
Presentation and
Frontend Development and Deployment
finishing the project
USER
INTERFACE
USER
INTERFACE
USER
INTERFACE
USER
INTERFACE
USER
INTERFACE
CONCLUSION

Our project on diabetic retinopathy detection using CNNs offers a


comprehensive approach to leveraging deep learning for early
diagnosis, emphasizing the potential for improving patient outcomes
through efficient and accurate screening methods. By harnessing the
power of convolutional neural networks, we aim to contribute to the
advancement of medical image analysis, paving the way for more
accessible and effective healthcare solutions in the field of diabetic
retinopathy management.
References used

https://www.nature.com/articles/s41467-021-
23458-5
https://link.springer.com/article/10.1007/s00521-023-
08249-x

https://pubmed.ncbi.nlm.nih.gov/36467440/

https://colab.research.google.com/drive/1UFA-
AvXJUnYakDEA5AQMj0eQReD_QM1ausp=sharing#scrollTo=yIjBpfNZa4YG
THANK YOU

You might also like