Diabetic Retinopathy Detection Using Amalgamated Deep Learning Algorithm

Document Type

Conference Article

Publication Title

Lecture Notes in Networks and Systems

Abstract

The eye disorder termed diabetic retinopathy (DR), which can cause diminished vision, can be brought on by diabetes. DR detection and routine diagnosis are complex tasks that may require multiple testing. Vision loss may be avoided or delayed with early detection of DR. However, the early diagnosis of DR is a challenging mission that necessitates the interpretation of fundus images by clinical specialists. Deep Learning Models (DLM) have become effective techniques for medical image analysis in recent years, promising to provide precise and automated DR identification. DLM automatically extracts the most discriminative features from training photos, but which characteristics are removed to produce predictions is unknown. This paper provides a blended DL method for categorizing DR in fundus pictures. To reduce over-fitting brought on by imbalanced datasets within a single DLN, our strategy employs several of the most well-liked DLN methods learned and validated in a balanced image set, merging their findings in a composite framework. The approach enhances robustness by reducing potential over-fitting patterns and produces more reliable predicted outputs than those obtained using individual DL designs. It does this using the advantages of developing the DLN in multiple resolutions. The suggested method successfully classifies glaucoma images with a sensitivity of 92% and a specificity of 93%.

First Page

100

Last Page

111

DOI

10.1007/978-3-031-55848-1_12

Publication Date

1-1-2024

Link to Full Text

Share

COinS