Glaucoma classification based on intra-class and extra-class discriminative correlation and consensus ensemble...

Area of Science:

• Ophthalmology
• Medical Imaging
• Machine Learning

Background:

• Glaucoma diagnosis from fundus images is crucial for Computer-Aided Diagnosis (CAD) systems.
• Accurate and efficient automated classification methods are needed to aid clinicians.

Purpose of the Study:

• To propose a novel fused feature extraction and ensemble classifier fusion technique for glaucoma diagnosis.
• To enhance the accuracy and convergence of glaucoma classification using fundus images.

Main Methods:

• A three-stage approach involving image preprocessing, feature extraction, and feature fusion using Intra-Class and Extra-Class Discriminative Correlation Analysis (IEDCA).
• Ensemble classification using Support Vector Machine (SVM), Random Forest (RF), and K-Nearest Neighbor (KNN), followed by classifier fusion via Consensus-based Combining Method (CCM).

Main Results:

• The proposed method achieved a classification accuracy of 99.2% on public datasets (HRF and DRIVE).
• The fusion classifier demonstrated improved accuracy and convergence compared to individual classifiers.
• Cross-dataset validation confirmed the robustness of the approach.

Conclusions:

• The novel fused feature extraction and ensemble classifier fusion technique offers a highly accurate and effective method for glaucoma diagnosis.
• This approach shows significant potential for integration into Computer-Aided Diagnosis (CAD) systems for ophthalmology.