Detection of Glaucoma from Fundus Images Using Novel Evolutionary-Based Deep Neural Network

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Glaucoma is an asymptomatic optic nerve condition often caused by high intraocular pressure, leading to irreversible blindness.
• Early glaucoma detection is crucial for treatment and preventing vision loss, but accurate diagnosis remains challenging.
• Current diagnostic methods require improvement for timely intervention.

Purpose of the Study:

• To propose a novel glaucoma detection system for accurate diagnosis of glaucoma from retinal images.
• To develop an effective classification model for distinguishing between healthy and glaucoma-affected eyes.
• To enhance early detection rates for glaucoma to prevent permanent vision impairment.

Main Methods:

• A four-phase approach: data preprocessing/enhancement, segmentation, feature extraction, and classification.
• Development of a novel fractional gravitational search-based hybrid deep neural network (FGSA-HDNN) classifier.
• Utilizing specific characteristics of optic nerve images for diagnosis.

Main Results:

• The proposed FGSA-HDNN model achieved a high diagnostic accuracy of 98.75% for glaucoma detection.
• Experimental analysis demonstrated superior performance compared to various existing techniques.
• The system effectively classified glaucoma-infected images from normal ones.

Conclusions:

• The developed glaucoma detection system, utilizing FGSA-HDNN, shows significant promise for accurate and early diagnosis.
• The proposed method offers a reliable tool for ophthalmologists in identifying glaucoma.
• High accuracy rates suggest potential for clinical application in preventing blindness.