TITLE:
Identification of Diabetic Retinal Exudates in Digital Color Images Using Support Vector Machine
AUTHORS:
R. F. Mansour, E. Md. Abdelrahim, Amna S. Al-Johani
KEYWORDS:
Hard Exudates; Diabetic Retinal; Support Vector Machine; Discrete Cosine Transform
JOURNAL NAME:
Journal of Intelligent Learning Systems and Applications,
Vol.5 No.3,
July
31,
2013
ABSTRACT:
Support vector machine (SVM) has become an increasingly popular tool for machine learning tasks involving classification. In this paper, we present a simple and effective method of detect and classify hard exudates. Automatic detection of hard exudates from retinal images is worth-studying problem since hard exudates are associated with diabetic retinopathy and have been found to be one of the most prevalent earliest signs of retinopathy. The algorithm is based on Discrete Cosine Transform (DCT) analysis and SVM makes use of color information to perform the classification of retinal exudates. We prospectively assessed the algorithm performance using a database containing 1200 retinal images with variable color, brightness, and quality. Results of the proposed system can achieve a diagnostic accuracy with 97.0% sensitivity and 98.7% specificity for the identification of images containing any evidence of retinopathy.