A Novel Context Aware Joint Segmentation and Classification Framework for Glaucoma Detection.

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Glaucoma is a leading cause of irreversible visual loss due to optic nerve damage.
• Early detection and intervention are crucial for managing glaucoma.
• Glaucoma diagnosis is challenging due to its subtle onset and patient variability.

Purpose of the Study:

• To propose a novel deep learning framework, GD-YNet, for optic disc segmentation and glaucoma detection.
• To leverage aggregated transformations and YNet architecture for context-aware analysis.
• To achieve high accuracy in binary classification for glaucoma detection.

Main Methods:

• Developed a context-aware deep learning framework named GD-YNet.
• Utilized aggregated transformations and YNet architecture for optic disc segmentation.
• Trained and validated the model on RIGA and RIMOne-V2 datasets.

Main Results:

• Achieved high glaucoma detection accuracies across multiple datasets: 99.72% (ACRIMA), 98.02% (Drishti-gs), 99.50% (REFUGE), and 99.41% (RIMOne-V1).
• Demonstrated robust performance in optic disc segmentation.
• The model shows potential for extension to multi-class classification for staging and severity assessment.

Conclusions:

• The GD-YNet framework provides a highly accurate and robust method for glaucoma detection.
• Deep learning approaches, like GD-YNet, show significant promise for improving glaucoma diagnosis.
• The model's adaptability suggests future applications in glaucoma staging and severity assessment.