Glaucoma Detection Using Image Processing and Supervised Learning for Classification.

Area of Science:

• Biomedical Engineering
• Ophthalmology
• Computer Science

Background:

• Manual detection of physiological changes is challenging, time-consuming, and prone to errors.
• Early disease detection is crucial for effective treatment and management.
• Computer-assisted diagnostics (CAD) offer a promising solution for improving diagnostic accuracy and efficiency.

Purpose of the Study:

• To develop a computer-aided design (CAD) system for the early identification, screening, and treatment of glaucoma.
• To evaluate the potential of an image analysis model for diagnosing glaucoma and other ocular disorders.
• To provide ophthalmologists with a decision support tool for diagnosing eye diseases.

Main Methods:

• An ensemble-based deep learning model was developed for glaucoma diagnosis.
• Three pre-trained convolutional neural networks (ResNet, VGGNet, GoogLeNet) were utilized.
• The model was trained and validated on five diverse datasets: DRISHTI-GS, DRIONS-DB, HRF, and PSGIMSR.

Main Results:

• The proposed ensemble architecture achieved 91.11% accuracy on the PSGIMSR dataset.
• Sensitivity of 85.55% and specificity of 95.20% were recorded on the PSGIMSR dataset.
• High accuracy rates were also observed on other datasets, including 95.63% (DRIONS-DB), 98.67% (HRF), and 95.64% (DRISHTI-GS).

Conclusions:

• The developed CAD system demonstrates significant potential for the early and accurate diagnosis of glaucoma.
• The ensemble deep learning approach offers a reliable and efficient method for ocular disorder evaluation.
• This system can serve as a valuable assistive tool for ophthalmologists, enhancing diagnostic capabilities.