Conference Paper
Automated Early Detection and Classification of Diabetic
Retinopathy Stages Using Deep Learning on Fundus Images
Tokare, K.N.*1, Siddiqui, I.R.1, Memon, A.R.1
1M.H. Saboo Siddik College of Engineering, Byculla, Mumbai 400008, India.
September 2024
Notice: This manuscript has been authored by officials of M.H. Saboo Siddik College of Engineering and submitted by the Final
Year Students of Computer Science Engineering (Specialization in Artificial Intelligence and Machine Learning).
This preprint is intended for publication in a journal or proceedings. Since changes may be made before publication, it may not be
cited or reproduced without the author’s permission.
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
Conference Paper
DISCLAIMER
This paper stems from the collaborative efforts of students at M.H. Saboo Siddik College of
Engineering. The research and findings shared here represent the hard work and dedication of the
authors.
It’s important to note that neither M.H. Saboo Siddik College of Engineering nor any of its
professors, administrators, or associated organizations can guarantee the accuracy, completeness,
or dependability of the information, techniques, products, or processes mentioned in this paper.
The authors also do not claim that any of the information or methods discussed will be free from
any potential conflicts with privately owned rights. Mentioning specific commercial products,
processes, or services—by trademark, trade name, manufacturer, or in any other way—doesn't
imply endorsement or recommendation by M.H. Saboo Siddik College of Engineering or its
representatives.
The insights, interpretations, and opinions expressed in this paper are solely those of the authors.
They do not necessarily reflect the views or policies of the college or its faculty. Any conclusions
drawn from this work are the authors’ own and shouldn’t be interpreted as representing the
institution or its affiliated entities.
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
Conference Paper
Automated Early Detection and Classification of Diabetic Retinopathy Stages Using
Deep Learning on Fundus Images
Tokare, K.N*1, Siddiqui, I.R1, Memon, A.R1
1
M.H. Saboo Siddik College of Engineering, Byculla, Mumbai 400008, India.
Abstract
Diabetic retinopathy (DR) stands as one of the primary causes of blindness around the world,
particularly impacting those who have struggled with elevated blood sugar levels over extended
periods[1] . For individuals living with diabetes, understanding and managing their condition is
crucial, and that includes keeping an eye on complications like DR. The key to preserving vision
lies in early detection, which can significantly reduce the risk of severe vision loss.
In this paper, we introduce an innovative approach to combat this pressing health issue by
proposing a deep learning-based model aimed at automating the detection and classification of
diabetic retinopathy stages using fundus images. Traditionally, diagnosing DR requires
specialized medical training and can be time-consuming; however, with advancements in
technology, we can now turn to powerful tools like convolutional neural networks (CNNs). These
sophisticated algorithms analyse fundus images—pictures of the interior surface of the eye—to
classify the stages of diabetic retinopathy into four distinct categories: mild, moderate, severe,
and proliferative.[2]
Our model demonstrates impressive accuracy, making it a promising tool for use in primary care
settings. This means that healthcare providers could quickly screen patients for diabetic
retinopathy during routine check-ups, allowing for timely diagnosis and intervention. By
integrating such a system into everyday healthcare practices, we hope to empower more medical
professionals to identify at-risk patients early, leading to better treatment options and ultimately
protecting the vision of countless individuals affected by diabetes. This approach could not only
alleviate the burden on specialists but also enhance the overall quality of care for those living
with this chronic condition, paving the way for improved health outcomes and quality of life.
KEYWORDS : Diabetic Retinopathy (DR), Early Detection, Deep Learning, Convolution
Neural Networks (CNNs), Fundus Images.
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
Conference Paper
1. Introduction
Diabetic retinopathy is a serious complication of diabetes characterized by damage to the retinal
blood vessels due to prolonged elevated blood sugar levels [1] .The progression of this condition
typically unfolds through four distinct stages: mild non-proliferative, moderate non-proliferative,
severe non-proliferative, and proliferative diabetic retinopathy (PDR) [2] . Early detection is vital
for effective management and prevention of vision loss associated with this disease. Currently,
comprehensive eye examinations are employed for diagnosis, which include visual acuity
assessments, dilated eye examinations, and advanced imaging techniques such as optical
coherence tomography (OCT) and fluorescein angiography [3]
Recent advancements in artificial intelligence (AI) have revolutionized the detection of diabetic
retinopathy, particularly through the analysis of fundus images. AI-based systems, including
• Google’s DeepMind [4]
• IDx-DR [5]
• EyeArt [6]
• RetinaLyze [7]
The above systems have shown remarkable sensitivity and specificity in identifying diabetic
retinopathy. These technological innovations not only enhance diagnostic accuracy but also hold
the potential to streamline the screening process, making it more accessible to patients. The
integration of AI into clinical practice represents a significant step forward in the fight against
diabetic retinopathy, emphasizing the importance of early intervention and ongoing monitoring
to preserve vision in affected individuals.
Diabetic retinopathy can be detected through a comprehensive eye exam, which may include:
1. Visual acuity test : Measures how well you see at various distances. [8]
2. Dilated eye exam : Drops are placed in your eyes to widen (dilate) the pupils, allowing
the doctor to examine the retina and optic nerve. [9]
3. Optical coherence tomography (OCT) : Imaging test providing cross-sectional images
of the retina. [10]
4. Fluorescein angiography : Dye is injected into your arm, and photos of the retina are
taken to show blood flow and detect leaks or abnormalities. [11]
• Severity:
Diabetic retinopathy is a leading cause of blindness in adults. The severity can range from mild,
with no symptoms, to severe vision loss and blindness if left untreated. [12]
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
Conference Paper
2. SOFTWARE and AI in Detection:
Several attempts have been made to create software and use artificial intelligence (AI) to aid in the
detection of diabetic retinopathy:
1. Google's DeepMind : Developed an AI system to detect over 50 eye diseases, including
diabetic retinopathy, with accuracy comparable to that of expert ophthalmologists. (The AI
system developed by DeepMind has shown performance comparable to that of expert
ophthalmologists. In tests, it demonstrated high sensitivity and specificity in detecting diabetic
retinopathy, making it a reliable tool for screening.)
2. IDx-DR : An FDA-approved AI-based software that can detect diabetic retinopathy in
primary care settings without the need for an eye specialist (The FDA approval was based on a
pivotal clinical trial involving 900 patients at 10 primary care sites. The trial demonstrated that
IDx-DR had a sensitivity (ability to correctly identify those with the disease) of 87.4% and a
specificity (ability to correctly identify those without the disease) of 89.5%.)
3. EyeArt : Another AI system approved by the FDA for the detection of diabetic
retinopathy.[6] (EyeArt has been validated through extensive clinical studies. In a pivotal trial
involving over 900 patients, EyeArt demonstrated sensitivity (ability to correctly identify those
with the disease) of 95.5% and specificity (ability to correctly identify those without the disease)
of 86.0%. These results indicate that EyeArt is highly effective in detecting diabetic retinopathy.)
4. RetinaLyze : An AI-based software used in Europe for early detection and monitoring of
diabetic retinopathy. ( RetinaLyze has undergone clinical validation to ensure its accuracy and
reliability. Studies have demonstrated its effectiveness in detecting diabetic retinopathy with high
sensitivity and specificity. These validations have supported its use as a dependable tool for
diabetic retinopathy screening.)
Fig.0.1. Pathologies [7]
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
Conference Paper
Fig.0.2. Final Retinopathy Severity Scale (For Individual Eyes) [ Link to pdf ]
Treatment :
• Control of Blood Sugar, Blood Pressure, and Cholesterol: Essential to prevent progression.
• Laser Surgery: Used to shrink or seal off leaking blood vessels. Link
• Vitrectomy:Surgery to remove blood from the vitreous and scar tissue from the retinaLink
• Anti-VEGF Injections: Medications injected into the eye to reduce swelling and slow the
growth of abnormal blood vessels. Link
3. HARDWARE for the Detection:
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
Conference Paper
Here's a more relatable overview of the hardware technologies used for detecting Diabetic
Retinopathy (DR), along with their price ranges, availability, and feasibility, presented in a reader-
friendly manner:
1. Fundus Cameras
Fundus cameras are sophisticated devices that take detailed photos of the back of the eye, including
the retina. They're essential tools in assessing eye health.
Fig.0.3. Fundus Camera
Prices:
- Basic models: $5,000 - $10,000
- Advanced models with automated features: $20,000 - $50,000
Availability: Widely found through medical suppliers and well-known brands like Canon, Topcon,
and Zeiss.
Feasibility:
- Pros: They provide high-quality images and are a staple in clinical environments, helping doctors
closely examine retinal structures.
- Cons: The initial investment can be steep, and trained personnel are necessary to operate these
machines and interpret the images successfully.
2. Optical Coherence Tomography (OCT)
OCT is a non-invasive method that gives cross-sectional images of the retina, enabling healthcare
professionals to see different retinal layers and identify any abnormalities.
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
Conference Paper
Fig.0.4. Optical Coherence Tomography (OCT)
Prices:
- Basic models: $15,000 - $30,000
- Advanced models: $50,000 - $100,000
Availability: Popular among manufacturers such as Heidelberg Engineering and Optovue.
Feasibility:
- Pros: Offers high-resolution imaging, capable of detecting subtle changes in the retina early on
and is useful for diagnosing a range of retinal diseases.
- Cons: The expense can be prohibitive, and skilled technicians are needed for both operation and
detailed interpretation of the images.
3. Automated Retinal Imaging Systems
These advanced systems use cutting-edge algorithms alongside imaging technology to quickly
capture and analyze fundus images for the automated detection of Diabetic Retinopathy.
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
Conference Paper
Fig.0.5. Automated Retinal Imaging Systems
Prices: $20,000 - $50,000 for systems that come equipped with AI capabilities.
Availability: Offered by innovation-driven companies like IDx Technologies[5] and EyeArt[6].
Feasibility:
- Pros: They streamline the screening process, making it quicker and easier for healthcare
providers, and can be effectively used in primary care settings.
- Cons: Although efficient, the upfront cost remains a consideration, and incorporating these
systems into existing healthcare frameworks might require additional funding.
4. Handheld Fundus Cameras
These portable devices make it easy to take images of the retina, particularly in remote or
underserved areas, offering flexibility in eye care.
Fig.0.6. Handheld Fundus Camera
Prices: Typically range from $1,000 to $10,000, which varies depending on the model and features.
Availability: Available from manufacturers like Welch Allyn and Optomed.
Feasibility:
- Pros: Their portability and lower cost make them attractive for use in various settings,
including community clinics.
- Cons: They may produce lower image quality compared to standard stationary fundus
cameras, and additional training might be necessary for optimal use.
[ Each of these technologies plays a vital role in the early detection and management of Diabetic
Retinopathy, helping to safeguard eye health in various clinical settings.]
Hardware ( RetinaLens )
An affordable device for Early Detection of Diabetic Retinopathy
Even with the improvements in technology for detecting diabetic retinopathy (DR), we still face
some important challenges. Traditional equipment like fundus cameras and optical coherence
tomography (OCT) can be very expensive, costing anywhere from $5,000 to $100,000. This high
price tag limits their use in primary care settings and underserved communities. Additionally, these
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
Conference Paper
devices require trained professionals to operate and interpret the results, which can slow down the
diagnosis process.
Automated retinal imaging systems offer quick results, but the initial costs and difficulties
integrating them into existing healthcare practices make them less practical for smaller clinics.
Handheld fundus cameras are a more affordable option, but they often sacrifice image quality, and
users still need some training to use them effectively.
To overcome these obstacles, I'm working on an innovative solution that will make early diabetic
retinopathy detection more accessible. The device will use a 20 Diopter lens connected to hardware
that lets users take fundus images with their smartphone cameras. We expect this device to cost
between $47 and $72, making it far more affordable than current options.
The images captured will be analyzed using specialized software designed to accurately classify
the stages of DR. This approach aims to provide wider access to detection methods, empowering
both healthcare providers and patients to catch the disease early and intervene sooner. Ultimately,
this could lead to better health outcomes and help preserve vision for those at risk.
Fig.0.7. 20 Diopter Lens (Attached to our Hardware with a Smartphone)
The RetinaLens is an innovative device that makes it easy for people to check for early signs of
diabetic retinopathy right from their homes. Its user-friendly design means that anyone can quickly
learn how to take fundus images without needing extensive training.
With just a little practice, even those without medical backgrounds can feel confident using the
device to keep an eye on their eye health. This accessibility empowers individuals living with
diabetes to take control of their health and catch potential issues early on. By simplifying the
process and making it more approachable, the RetinaLens aims to ensure that everyone could
monitor their eye health, ultimately leading to better health outcomes for all.
Software ( RetinaLens )
An easy-to-use Software for Early Detection of Diabetic Retinopathy
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
Conference Paper
Fig.0.8. Diabetic Retina
The software which we are building takes advantage of the latest deep learning advancements,
using YOLOv8[21], a powerful tool known for its ability to detect objects in real-time. We’ve
trained it using a vast collection of fundus images from Kaggle[22], carefully organized by the
different stages of diabetic retinopathy—mild, moderate, severe, and proliferative. By neatly
categorizing these images, we’ve made it easier for the model to learn how to spot the different
stages of the disease, ensuring it becomes as precise and accurate as possible.
Data Preparation & Feature Extraction:
To get the best results from YOLOv8, we put a lot of effort into preparing the images. This includes
improving them by normalizing brightness, resizing, and enhancing quality so that they’re
consistent and clear. The most important features—like tiny blood vessel damage, haemorrhages,
and abnormal growths—are highlighted, making it easier for the model to recognize them. This
helps the model focus on what really matters and allows it to differentiate between each stage of
the disease with accuracy.
Model Training & Optimization:
Once the data is ready, we train the YOLOv8 model to recognize diabetic retinopathy stages with
the highest accuracy possible. YOLOv8 is great for medical imaging because it can analyse the
entire image quickly and efficiently. It doesn’t just process the image— it detects subtle early signs
of diabetic retinopathy, making it a valuable resource for both doctors and patients, allowing for
early intervention when it matters most.
Application:
We’ve designed the software to be easy to use, whether in a clinical setting or at home. Doctors
can quickly and reliably screen patients for diabetic retinopathy during routine exams, and with
our affordable RetinaLens, individuals can even monitor their eye health at home. This makes
frequent check-ups possible, especially for people in remote areas where seeing an eye specialist
might not be convenient. Our system empowers people to take control of their health and catch
problems before they become severe.
Key Features:
• Real-time Detection: The YOLOv8 model processes fundus images instantly, giving users
fast and accurate results on the spot.
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
Conference Paper
• Stage Classification: It accurately identifies which stage of diabetic retinopathy—mild,
moderate, severe, or proliferative—the patient is in, helping track the progression of the
disease.
• Ease of Use: The software is simple to use, even for non-experts, making it accessible to
both healthcare providers and individuals.
• High Accuracy: Thanks to rigorous training and continuous improvement, our software
ensures highly accurate detection of diabetic retinopathy.
• Affordable Solution: Paired with our cost-effective hardware, the system provides an
affordable alternative to traditional DR screening tools, making it accessible to a much
broader audience.
Fig.0.9. Confusion Matric for YoloV8 on detection of DR
Accuracy and Performance:
We’re committed to achieving the highest level of accuracy. That’s why we continually test, adjust,
and improve the model through multiple rounds of training. Our goal is to minimize errors and
ensure it can detect both early and advanced stages of diabetic retinopathy with precision. By
thoroughly testing across different patient groups, we ensure that the model performs consistently,
matching the accuracy of trained eye specialists.
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
Conference Paper
Fig.1.0. Stages detected by YoloV8
Vision for the Future:
Our mission is to make diabetic retinopathy detection more affordable and accessible to everyone,
especially those who may not have regular access to eye care. By combining advanced deep
learning technology with budget-friendly hardware, we aim to close the gap between early
detection and timely treatment. Ultimately, we hope to reduce the global burden of diabetic
retinopathy and help prevent blindness in those affected by this condition.
In the future, we’ll continue enhancing the software, adding features that can predict how the
disease might progress and offer treatment recommendations. This will not only improve early
detection but also empower healthcare providers with a tool that helps them manage diabetic
retinopathy more effectively, leading to better outcomes for patients everywhere.
Fig .1.1. Comparing different Deep Learning model for Detection
3. Results
The below figure shows the representation of our android application based on Flutter named
as ‘RetinaLens’ that would take input fundus images from the user and then the application
will process the image and categories the image Diabetic Retinopathy found or not and further
will segment into the Stages of Diabetic Retinopathy.
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
Conference Paper
Fig.1.2. Figma Design of RetinaLens (Android Application for early detection of
Diabetic Retinopathy)
This combination of Hardware and Software would allow the user to :
1. Spot Eye Problems Early
Imagine having a tool that uses advanced retinal imaging to capture clear pictures of the inside of
your eye. This technology makes it possible to catch issues like diabetic retinopathy, glaucoma,
and macular degeneration before they lead to serious vision loss. The best part? It’s paired with
smart AI software that can analyse these images and highlight any early signs of disease, which is
crucial since treating these conditions early can save your sight.
2. Making Eye Care Accessible
Think about how portable and affordable retinal scanners can bring eye care to remote areas where
specialized services might not exist. By using these devices, healthcare providers can perform eye
screenings for more people, no matter where they live. Plus, with apps and cloud technology like
RetinaLens, they can easily share retinal images with specialists around the world, ensuring that
even those in underserved regions can receive expert feedback without needing to travel.
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
Conference Paper
3. Saving Time and Money
With compact and user-friendly devices, eye screening can be quick and efficient, especially in
busy clinics or during community health events. AI-driven tools can process the results
automatically, saving doctors valuable time and reducing the amount of manual work required to
evaluate large numbers of patients. This efficiency not only makes care faster but also helps to
lower costs for both providers and patients.
4. Better Accuracy in Diagnosis
Advanced imaging technology provides high-resolution images that are essential for accurately
diagnosing eye conditions. AI can sift through vast amounts of data, spotting tiny details or
irregularities that might escape even the most trained eyes. By minimizing human error, these tools
significantly enhance diagnostic accuracy.
5. Effortless Data Management
The software involved often includes features for storing retinal images and tracking patient
histories over time. This makes it simple to monitor how a disease progresses, compare past scans,
and evaluate treatment effectiveness. Cloud storage ensures that this data can be securely shared
among different specialists.
6. Scaling Up and Remote Care
Once the retinal imaging tool is set up, it can be used in various locations like hospitals, clinics, or
even mobile health units, making it adaptable to different healthcare environments. Remote
monitoring and telemedicine capabilities enable healthcare providers to review eye scans from
anywhere, ensuring that patients at high risk receive continuous care without needing frequent
inperson visits.
7. Encouraging Proactive Eye Health
With regular access to eye screenings, patients are more likely to take charge of their eye health.
These early detection tools not only raise awareness and encourage routine check-ups but also
promote timely interventions before conditions worsen.
8. Boosting Research and Development
The software can continue to learn and improve, thanks to AI and machine learning, by analyzing
real-world data. This means better predictive models for eye diseases over time. Additionally, the
tool can contribute to clinical research by providing large, anonymized datasets, which are
invaluable for investigating various eye conditions.
9. User-Friendly Convenience
The interface of the retinal imaging device is designed to be intuitive, so even non-specialists like
technicians or general practitioners can use it easily. Mobile apps or web interfaces, like
RetinaLens, facilitate smooth interactions with the device whether you’re a healthcare professional
or a patient, enhancing overall user experience and convenience.
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
Conference Paper
10. Fighting Blindness and Vision Loss
Ultimately, these innovative tools can play a significant role in reducing the number of people who
suffer from avoidable blindness and vision loss. By offering early, precise, and accessible
diagnoses, they improve the quality of life for countless individuals around the world.
4. Difficulties in Hardware & Software :
• Hardware Development Challenges
1. Precision Imaging: Capturing high-quality, detailed images of the retina is essential for
accurately detecting diabetic retinopathy. One major challenge is developing an affordable device,
like the RetinaLens, that can produce clear images using a smartphone. This involves ensuring that
the lens offers clarity, maintains focus, and provides adequate lighting in various environments,
which can be quite complex.
2. Cost vs. Quality: Our goal is to create a low-cost device priced between $47 and $72, but
achieving this while maintaining image quality comparable to expensive fundus cameras is a
difficult balancing act. We constantly navigate the trade-offs between affordability and
performance.
3. User Accessibility: The hardware needs to be user-friendly, especially for non-specialists.
Designing a device that is both effective and intuitive for patients to use at home requires extensive
testing and adjustments based on user feedback.
4. Integration with Smartphones: Making sure the device is compatible with a wide range of
smartphone models and operating systems adds another layer of complexity. It’s important that the
device works seamlessly with different camera qualities and hardware specifications.
5. Durability and Portability: We need to ensure that the device is durable and portable while
maintaining its performance over time and in various conditions. Achieving this at a low cost poses
significant challenges.
• Software Development Challenges
1. Training Deep Learning Models: Creating an accurate model, such as YOLOv8, to detect
diabetic retinopathy requires extensive training on a large and diverse dataset. Collecting and
organizing a high-quality dataset that includes enough examples of each stage of diabetic
retinopathy is both time-consuming and resource-intensive.
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
Conference Paper
2. Feature Extraction: Accurately identifying subtle features in the images, such as
microaneurysms, haemorrhages, and abnormal blood vessels, is a significant challenge. We need
to extract these features in a way that allows the model to learn and generalize effectively.
3. Ensuring High Accuracy: It’s crucial to achieve the highest possible accuracy while
minimizing false positives and false negatives, especially in a medical context. The model must
undergo rigorous testing and optimization to perform at or above the level of trained
ophthalmologists.
4. Cross-device Compatibility: The software must run efficiently on various platforms, from
smartphones to medical computers. Ensuring smooth operation across different devices, operating
systems, and screen sizes adds to the complexity of development.
5. Real-time Processing: Since YOLOv8 processes images in real-time, the software needs to
be optimized for quick results without sacrificing accuracy. Striking the right balance between
speed and precision requires ongoing refinement.
6. Security and Privacy: Handling medical data means we must comply with strict security
and privacy regulations, such as HIPAA. Implementing robust encryption and secure data storage
protocols adds another layer of complexity to the development process.
5. Conclusions
In this paper, we present a comprehensive solution for the Early detection and classification
diabetic retinopathy (DR) using YOLOv8, a powerful deep learning model, along with an
affordable hardware device we’ve named the RetinaLens. Our model has been trained on a large
dataset from Kaggle, focusing on important retinal features such as microaneurysms and
hemorrhages, enabling us to classify DR into four distinct stages. The RetinaLens allows users
to capture high-quality fundus images directly with their smartphones, making DR screening both
accessible and affordable, with a price range of $47 to $72.
During the development process, we tackled several challenges, including ensuring the quality of
the images, maintaining the accuracy of our model, and striking a balance between affordability
and effectiveness. By integrating this hardware with our software, which processes images in
real-time, we aim to deliver a reliable and user-friendly solution that can be used in both clinical
settings and at home. This system empowers individuals to take charge of their eye health,
particularly in underserved areas where access to specialists is limited.
Looking ahead, we plan to test additional models alongside YOLOv8 to further improve detection
accuracy. Our goal is to provide a robust, affordable, and accessible solution for diabetic
retinopathy screening that can help reduce the risk of blindness and enhance healthcare outcomes
for diabetic patients around the world.
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
Conference Paper
References
1) Mohammad Irfan Dar a b, Anu Gulya c, Hallmarks of diabetes mellitus and insights into
the therapeutic potential of synergy-based combinations of phytochemicals in reducing
oxidative stress-induced diabetic complications, Natural Product Research,
https://www.tandfonline.com/doi/full/10.1080/14786419.2024.2402461, September 2024.
2) Shivappriya S N, Alagumeenaakshi M, Sasikala S, Deep Learning-Based Diabetic
Retinopathy Severity Classification and Progression Time Estimation, 22-IFAC
conference, https://www.sciencedirect.com/science/article/pii/S240589632400212X, May
2024.
3) Jacqueline Chua , Bingyao Tan Optical coherence tomography angiography of the retina
and choroid in systemic diseases, Progress in retinal and eye research Vol-103, Science
Direct, https://www.sciencedirect.com/science/article/pii/S1350946224000570,
November 2024.
4) Google DeepMind, https://deepmind.google/, Company Founded in UK in November
2010. 5) IDx-Dr-Healthviser, https://www.healthvisors.com/en/idx-dr/ , Founded in Spain
in 2017.
6) EyeArt, Kaushal Solanki, Paul Praino, Chaithanya Ramachandra,
https://www.eyenuk.com/en/products/eyeart/, Founded in 2010.
7) Retinalyze, https://www.retinalyzebrasil.com.br/en/, Cona Health Care Brasil LTDA,
Founded in 2018.
MHSSCOE @all rights reserved, Conference Paper, Major Project, Diabetic Retinopathy (DR)
