Sparse dissimilarity-constrained coding for glaucoma screening.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Glaucoma is a leading cause of irreversible blindness.
• Early detection and treatment are vital for slowing disease progression.
• Current population-based screening methods for glaucoma lack cost-effectiveness and accessibility.

Purpose of the Study:

• To propose an automated method for cup to disc ratio (CDR) assessment using 2-D retinal fundus images.
• To develop a low-cost, effective tool for glaucoma screening.
• To improve the accuracy of automated optic nerve head assessment.

Main Methods:

• Optic disc segmentation and reconstruction using a novel sparse dissimilarity-constrained coding (SDC) approach.
• SDC incorporates both dissimilarity and sparsity constraints from reference discs with known CDRs.
• CDR computation using reconstruction coefficients derived from the SDC method.

Main Results:

• The proposed method achieved an average CDR error of 0.064 and a correlation coefficient of 0.67 against manual measurements in a 650-image database.
• For glaucoma screening, the method demonstrated areas under the curve (AUC) of 0.83 and 0.88 on datasets of 650 and 1676 images, respectively.
• Performance surpassed existing state-of-the-art methods in both CDR assessment and glaucoma screening.

Conclusions:

• The developed method provides accurate CDR assessment for glaucoma detection.
• This automated approach shows significant potential for large-scale, population-based glaucoma screening initiatives.
• The findings support the use of 2-D retinal fundus images for efficient and cost-effective eye disease screening.