Automated Corneal Segmentation of Anterior Segment Photographed Images using Centroid-Based Active Contour Model

Abstract

In this paper, we propose an automatic Active Contour Model (ACM) using corneal centroid information for corneal segmentation. The algorithm was validated on 259 of Anterior Segment Photographed Images (ASPIs) that were captured using a smartphone. The proposed system comprises of four components; a) data collection, b) pre-processing, c) corneal segmentation using proposed automated centroid-based ACM, and (d) feature extraction. There were five corneal shape parameters employed, namely Horizontal Visible Iris Diameter (HVID), Vertical Visible Iris Diameter (VVID), eccentricity, asphericity and solidity. Each of these parameters has its own significant features in measuring the corneal shape quantitatively, which could potentially assist the ophthalmologists during a screening process. Using all ASPIs, the findings show that this method is competent in segmenting 231 corneal regions while Circular Hough Transform method only manage to segment 165 regions successfully. These segmentation results demonstrate a bright potential towards the development of intelligent classification system for ocular diseases.