Real-Time Glaucoma Detection From Digital Fundus Images Using Self-ONNs

Ozer Can Devecioglu,Serkan Kiranyaz,Turker Ince,Moncef Gabbouj,Junaid Malik,Eray Atalay

doi:10.1109/access.2021.3118102

Abstract

Glaucoma leads to permanent vision disability by damaging the optical nerve that transmits visual images to the brain. The fact that glaucoma does not show any symptoms as it progresses and cannot be stopped at the later stages, makes it critical to be diagnosed in its early stages. Although various deep learning models have been applied for detecting glaucoma from digital fundus images, due to the scarcity of labeled data, their generalization performance was limited along with high computational complexity and special hardware requirements. In this study, compact Self-Organized Operational Neural Networks (Self-ONNs) are proposed for early detection of glaucoma in fundus images and their performance is compared against the conventional (deep) Convolutional Neural Networks (CNNs) over three benchmark datasets: ACRIMA, RIM-ONE, and ESOGU. The experimental results demonstrate that Self-ONNs not only achieve superior detection performance but can also significantly reduce the computational complexity making it a potentially suitable network model for biomedical datasets especially when the data is scarce.

Highlights

Glaucoma, called “the silent thief of sight,” leads to permanent vision disability by damaging the optic nerve
As the first study where Self-Operational Neural Networks (ONNs) are evaluated against deep Convolutional Neural Networks (CNNs) over a classification problem, we have performed both fair and unfair comparisons
We have even used deep CNN models pre-trained and applied transfer learning on the benchmark datasets

Summary

INTRODUCTION

Called “the silent thief of sight,” leads to permanent vision disability by damaging the optic nerve. They applied pre-processing, morphological operations and thresholding for automatically detecting the OD, the blood vessels and computing the features These features are validated by classifying the normal and glaucoma fundus images collected at the Kasturba Medical College, India using a neural network classifier. This is why conventional homogenous networks such as CNNs fail to learn the problems whenever the solution space is highly nonlinear and complex [14]-[21] unless a sufficiently high network depth and complexity (variants of CNN) are accommodated To address these limitations, Operational Neural Networks (ONNs)[21]-[24] have recently been proposed as a heterogeneous network model encapsulating distinct nonlinear neurons. We propose compact Self-ONNs for glaucoma detection in fundus images and evaluate their performance extensively over the three benchmark datasets This is the first study where Self-ONNs are evaluated against deep CNNs over a classification problem.

THE PROPOSED APPROACH

Glaucoma Diagnosis Benchmark Datasets

CONCLUSIONS

Self-ONNs