Improving Diabetic Retinopathy Detection by using Image Processing and Machine Learning

Abstract

The rapid development and proliferation of medical imaging technologies is revolutionizing medicine. Medical imaging allows scientists and physicians to glean potentially life-saving information by peering noninvasively into the human body. Diabetic retinopathy (DR) is an irreversible fundus retinopathy. A deep learning-based auto-mated DR diagnosis system can save diagnostic time. This paper reviews and analyses state-of-the-art deep learning methods in supervised, self-supervised and Vision Transformer setups, proposing retinal fundus image classification and detection. For instance, referable, non-referable and proliferative classifications of Diabetic Retinopathy are reviewed and summarized. Moreover, this paper discusses the available retinal fundus datasets for Diabetic Retinopathy that are used for tasks such as detection, classification and segmentation. The paper also assesses research gaps in the area of DR detection/classification and addresses various challenges that need further study and investigation.

Introduction

Human Eye and Retina:
The human eye works like a camera, with the cornea and lens focusing light onto the retina, which contains photoreceptors—rods (for low-light vision) and cones (for color vision). The macula and its central fovea enable sharp, high-resolution vision. The optic disc is a blind spot where no photoreceptors exist. The retina’s transparency allows for non-invasive imaging, useful for diagnosing eye diseases (like macular degeneration, glaucoma) and systemic diseases (such as diabetes and hypertension).

Diabetic Retinopathy (DR):
DR is a diabetes-related complication causing retinal vessel damage, ischemia, and macular edema, leading to vision loss. Management involves blood sugar control, lifestyle changes, and treatments like laser therapy and anti-VEGF drugs. Early detection via regular screening is crucial.

Medical Image Processing:
Medical imaging (X-ray, CT, MRI, retinal imaging) is vital in diagnosis and treatment. Advances in digital image processing enable better reconstruction and analysis.

Proposed Work for DR Diagnosis:
A transformer-based deep learning system is proposed for efficient DR diagnosis using high-resolution retinal images. Techniques include:

• Segmenting images into patches for detailed analysis.

• Using Vision Transformer (ViT) models for classification into DR stages.

• Employing the HBA-U-Net architecture with hierarchical attention blocks for precise image segmentation.

• Developing a hybrid ISVM-RBF model combining improved SVM with other classifiers to enhance classification accuracy.

• Extracting features with CNNs and reducing dimensionality using FFT and Singular Value Decomposition (SVD) for efficient processing.

Classification Approach:
Images undergo preprocessing (noise removal, enhancement, resizing). A two-way cascade CNN extracts local and global features for classifying DR severity into four categories: none, mild, moderate, and severe.

Datasets Used:
The study uses large retinal image datasets (IDRiD and Kaggle DR Detection) annotated with clinical severity grades to train and validate the models.

Conclusion

This project presents an efficient method for detecting blood vessels and hard exudates in retinal fundus images, aiding in the early diagnosis of diabetic retinopathy. Image segmentation techniques were applied effectively, with performance influenced byfactors like intensity and texture. The system demonstrated reduced human error and suitability for remote diagnostics with low computational cost. It serves as a valuable screening tool for early DR detection by identifying key clinical features such as vessels, exudates, and optic disc. Future work should focus on precise exudate extraction using a hybrid of supervised and unsupervised methods. Combining CNNs with clustering techniques can enhance accuracy, reduce bias, and improve detection scalability. This approach could significantly advance diabetic retinopathy screening and broader medical imaging applications.

References

[1] V.Kumari and N.Suriyanarayanan, \"Feature Extraction for Early Detection of Diabetic Retinopathy,\" In International Conference on Recent trends in Information, Telecommunication and Computing, 2010, Pp. 359-361. [2] S.Ravishankar, \"Automated Feature Extraction for Early Detection of Diabetic Retinopathy in Fundus Images\" In Computer Vision and Pattern Recognition, 2009. CVPR 2009. IEEE Conference On 2009, Pp. 210-217. [3] Neeraj Sharma and Lalit M.Aggarwal, “Automated Medical Image Segmentation Techniques” Journal of Medical Physics, 2010 Jan-Mar [4] K.V.Kauppi T, Lensu L, Sorri I, Raninen A, \"Diaretdb1: Diabetic Retinopathy Database and Evaluation Protocol. In: Medical Image Understanding and Analysis (MIUA).” Ed, 2007. [5] Younis N, Broadbent Dm, Harding Sp, Vora Jp. Incidence Of Sight-Threatening Retinopathy in Type 1 Diabetes in a Systematic Screening Programme. Diabet Med ; 20:758-65.

Copyright

Copyright © 2025 H. Rajeswary, K. Murugan, B. Mohammed Umar, C. Shanmuga Prasath. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.