Obtaining optic disc center and pixel region by automatic thresholding methods on morphologically processed fundus...

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Automated retinal disease diagnosis systems require accurate identification of fundal landmarks like the optic disc (OD).
• Early detection of ophthalmic signs is crucial for timely intervention and management of retinal diseases, particularly in diabetic patients.

Purpose of the Study:

• To develop and evaluate an automated methodology for accurate center-position location and optic disc (OD) retinal region segmentation on digital fundus images.
• To assess the performance of the proposed method using publicly available datasets of diabetic patients' retinal images.

Main Methods:

• Iterative opening-closing morphological operations for region enhancement.
• A two-step automatic thresholding procedure to define a region of interest.
• Circular Hough transform combined with Prewitt edge detection for OD center and boundary identification.

Main Results:

• Highly accurate OD-center location with average Euclidean distances of 6.08, 9.22, and 9.72 pixels for different image sizes.
• Effective OD segmentation evaluated using Jaccard and Dice coefficients and mean average distance.
• The proposed method demonstrated superior overall performance compared to existing OD segmentation techniques.

Conclusions:

• The automated OD location and segmentation methodology is accurate and robust.
• This method is a suitable tool for integration into prescreening systems for early retinal disease diagnosis.
• The system holds significant potential for improving the efficiency of diabetic retinopathy screening programs.