Research Article
Multi-disease Classification of retinal images using Convolutional Neural Network
|
International Journal of Computer Applications
Foundation of Computer Science (FCS), NY, USA
|
| Volume 186 - Issue 31 |
| Published: July 2024 |
| Authors: N. Devi, P. Leela Rani, A.R. Guru Gokul |
10.5120/ijca2024923850
|
N. Devi, P. Leela Rani, A.R. Guru Gokul . Multi-disease Classification of retinal images using Convolutional Neural Network. International Journal of Computer Applications. 186, 31 (July 2024), 35-42. DOI=10.5120/ijca2024923850
@article{ 10.5120/ijca2024923850,
author = { N. Devi,P. Leela Rani,A.R. Guru Gokul },
title = { Multi-disease Classification of retinal images using Convolutional Neural Network },
journal = { International Journal of Computer Applications },
year = { 2024 },
volume = { 186 },
number = { 31 },
pages = { 35-42 },
doi = { 10.5120/ijca2024923850 },
publisher = { Foundation of Computer Science (FCS), NY, USA }
}
%0 Journal Article
%D 2024
%A N. Devi
%A P. Leela Rani
%A A.R. Guru Gokul
%T Multi-disease Classification of retinal images using Convolutional Neural Network%T
%J International Journal of Computer Applications
%V 186
%N 31
%P 35-42
%R 10.5120/ijca2024923850
%I Foundation of Computer Science (FCS), NY, USA
Abstract
In ophthalmology, early fundus screening is a cost-effective and efficient method to prevent blindness caused by eye diseases. Due to the lack of clinical evidence, manual detection is labor-intensive and can result in clinical delays. The advent of deep learning has shown promising results in diagnosing various eye diseases, though most studies focus on a single disease. Thus, a multi-disease classification approach using fundus images is highly effective. This paper introduces a method based on Convolutional Neural Networks (CNN) for classifying multiple diseases. The proposed method uses multi-scale ridge detection for segmentation and Dijkstra's algorithm to create a fully connected vascular tree. Typically, surgeons have angiographic data on hand and mentally register the images to pinpoint abnormalities. Superimposing angiographic edges onto the patient's retinal image accurately highlights the treatment area, making it easier to detect eye defects such as myopia, hyperopia, and diabetic retinopathy. The registered image is visually precise with high accuracy. The proposed model can classify five disease categories: age-related macular degeneration (ARMD), central retinal vein occlusion (CRVO), optic disc center (ODC), diabetic retinopathy (DR), and branch retinal vein occlusion (BRVO) with an overall accuracy of 92%.
References
No references available
Index Terms
Keywords