Document Type : Original Paper
Authors
1
M.Sc. Student of Medical Physics, Faculty of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran.
2
Assistant Professor, Medical Physics Dept., Faculty of Medicine, Shiraz University of Medical Sciences, Shiraz, Iran.
3
Ph.D. Student of Computer Engineering, Ferdowsi University, Mashhad, Iran
4
Associate Professor, Ophthalmology Dept., Faculty of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran.
5
Professor, Medical Physics Dept., Faculty of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran.
6
Associate Professor, Computer Engineering Dept., Ferdowsi University, Mashhad, Iran.
Abstract
Introduction: Diabetic retinopathy (DR) is the single largest cause of sight loss and blindness in the working age population of Western countries; it is the most common cause of blindness in adults between 20 and 60 years of age. Early diagnosis of DR is critical for preventing vision loss so early detection of microaneurysms (MAs) as the first signs of DR is important. This paper addresses the automatic detection of MAs in fluorescein angiography fundus images, which plays a key role in computer assisted diagnosis of DR, a serious and frequent eye disease.
Material and Methods: The algorithm can be divided into three main steps. The first step or pre-processing was for background normalization and contrast enhancement of the image. The second step aimed at detecting landmarks, i.e., all patterns possibly corresponding to vessels and the optic nerve head, which was achieved using a local radon transform. Then, MAs were extracted, which were used in the final step to automatically classify candidates into real MA and other objects. A database of 120 fluorescein angiography fundus images was used to train and test the algorithm. The algorithm was compared to manually obtained gradings of those images.
Results: Sensitivity of diagnosis for DR was 94%, with specificity of 75%, and sensitivity of precise microaneurysm localization was 92%, at an average number of 8 false positives per image.
Discussion and Conclusion: Sensitivity and specificity of this algorithm make it one of the best methods in this field. Using local radon transform in this algorithm eliminates the noise sensitivity for microaneurysm detection in retinal image analysis.
Keywords
Main Subjects
)