Adaptive weighted locality-constrained sparse coding for glaucoma diagnosis

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Glaucoma is a leading cause of irreversible blindness.
• Accurate diagnosis relies on metrics like the vertical cup-to-disc ratio (CDR).
• 2D retinal fundus images are crucial for automated glaucoma detection.

Purpose of the Study:

• To introduce a novel sparse coding approach for glaucoma diagnosis.
• To enhance the robustness and effectiveness of glaucoma diagnosis using retinal images.
• To automate the estimation of the vertical cup-to-disc ratio (CDR) for improved diagnostic accuracy.

Main Methods:

• Developed adaptive weighted locality-constrained sparse coding (AWLCSC).
• Constructed a weighted matrix by adaptively fusing multiple distance measurements.
• Extracted and reconstructed disc images using AWLCSC for CDR estimation.

Main Results:

• The proposed AWLCSC method demonstrated superior performance compared to existing approaches.
• Evaluated on DRISHTI-GS1 and RIM-ONE r2 databases, showing significant improvements.
• Successfully automated the estimation of CDR from retinal fundus images.

Conclusions:

• AWLCSC provides a robust and effective method for glaucoma diagnosis.
• The approach offers automated CDR estimation, aiding in early detection.
• Outperforms state-of-the-art methods, paving the way for advanced glaucoma screening.