Diagnosis of Corneal Arcus Using Statistical Feature Extraction and Support Vector Machine

Abstract

Corneal arcus is a white ring or arc deposited in the corneal region of the human eye. This corneal abnormality is significantly associated with the lipid disorders and atherosclerosis. In this paper, we proposed a computer-aided diagnosis system to detect the corneal arcus. The proposed method detects the corneal arcus using the statistical features extracted from the iris region of the eye image. The iris region is segmented from the other regions of the eye image using circular Hough transform (CHT). In order to achieve the better classification results, a morphological operation-based specular reflection removal and colour transformation-based enhancement methods are also developed in this paper. The proposed method was implemented and evaluated using the abnormal eye images from our own database and normal eye images collected from UBIRIS.v1 database. Our database contains the eye images with different grades of corneal arcus abnormality. The performance of our method was evaluated using the confusion matrix-based metrics. In the training phase, our method achieved a classification accuracy of 1. In the testing phase, our method achieved a classification accuracy of 0.96 with a positive predictive value 0.9791 and negative predictive value 0.9423.