Integrating holistic and local deep features for glaucoma classification.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Automated glaucoma detection is crucial for early diagnosis and treatment.
• Image classification methods show promise but face challenges with limited data and optic disc misalignment.

Purpose of the Study:

• To develop a robust classification-based approach for automated glaucoma detection.
• To address limitations of insufficient sample size and optic disc shape variations.

Main Methods:

• Utilized deep convolutional networks pre-trained on large generic datasets for feature representation.
• Combined holistic and local features to improve robustness against misalignment.
• Evaluated performance on the Origa dataset.

Main Results:

• Achieved an Area Under the Receiver Operating Characteristic Curve (AUC) of 0.8384 on the Origa dataset.
• Demonstrated the effectiveness of the proposed classification-based approach.

Conclusions:

• The novel approach effectively detects glaucoma by leveraging deep learning and feature fusion.
• This method offers a promising solution for automated glaucoma diagnosis in retinal image analysis.