International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 05 | May 2025
p-ISSN: 2395-0072
www.irjet.net
CNN-based Architecture with Attention Mechanisms for Enhanced Diabetic Retinopathy Detection and Classification Meenal Katole1, Prof. Pramila M Chawan2 1MTech student, Dept of Computer Engineering and IT, VJTI college Mumbai, Maharashtra, India
2Associate Professor, Dept of Computer Engineering and IT, VJTI college Mumbai, Maharashtra, India
---------------------------------------------------------------------***--------------------------------------------------------------------Abstract - Diabetic retinopathy is one of the leading causes of visual loss, and early detection is critical to effective treatment.
This research discusses in detail the current deep learning practices, particularly convolutional neural networks and attention mechanisms that are applied to improve the detection and grading of DR. The study considers different algorithms and technologies employed in this area, assesses their performance, and examines the way attention mechanisms increase CNN performance. We conclude by highlighting key developments and identifying future research directions.
Key Words:
Diabetic Retinopathy, Convolutional Neural Networks (CNN), Attention Mechanisms, Medical Image Classification, Retinal Images.
1.INTRODUCTION 1.1 Diabetic Retinopathy Diabetic Retinopathy is a severe complication of diabetes, characterized by damage to blood vessels of the retina, progressive visual impairment, and potentially leading to blindness. Early detection of this complication is very important to prevent the DR from progressing to an advanced stage, such as NPDR and PDR. Traditionally, DR detection has been performed manually by analyzing the retinal images. It is a very time-consuming process and prone to human errors; therefore, it requires an automated detection system.
1.2 Convolutional Neural Networks (CNNs) Convolutional neural networks have become powerful tools for image classification, including medical imaging of diabetic retinopathy. They inherently learn the spatial hierarchies of features from images without requiring manual feature extraction. Thus, they are quite efficient in the diagnosis of DR from fundus images. However, CNNs sometimes fail to capture the subtle yet critical features of fundus images, which are very vital in the diagnosis of DR in its early stages limitation to their effectiveness.
1.3 Attention Mechanisms To such CNNs, attention mechanisms are utilized to guide the network on where in the image it should focus its attention. Spatial attention guides the model on significant regions of attention, usually with hemorrhages and exudates, whereas channel attention models support the model to emphasize feature maps relevant to further processes. Integrating an attention mechanism will improve the focus of the important areas of the CNN models, classifying with high precision, and better interpreting the results.
1.4 Algorithms for DR Detection Various approaches have, therefore, been developed for DR detection: These techniques depend on the features extracted manually and then perform the classification, typically using algorithms such as support vector machines or random forest. • CNN-based methods: DR detection performance has improved significantly due to architectures such as ResNet and Inception, which learn features directly from retinal images using CNNs. • Hybrid approaches: Some techniques involve multi-model architectures, such as combining CNNs with other models for data augmentation or attention mechanisms to enhance feature focus.purposes.
© 2025, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 823