Document Type : Original Paper
Associate Professor, Computer Engineering Dept., Ferdowsi University, Mashhad, Iran.
Professor, Medical Physics Dept., School of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran.
Assistant Professor, Center of Research in Medical physics and Biomedical Engineering, Shiraz University of Medical Sciences, Shiraz, Iran.
PhD Student, Computer Engineering Dept., Ferdowsi University, Mashhad, Iran.
MSc Student, Medical physics Dept., School of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran.
Introduction: Diabetic retinopathy (DR) is one of the most serious and most frequent eye diseases in the world and the most common cause of blindness in adults between 20 and 60 years of age. Following 15 years of diabetes, about 2% of the diabetic patients are blind and 10% suffer from vision impairment due to DR complications. This paper addresses the automatic detection of microaneurysms (MA) in color fundus images, which plays a key role in computer-assisted early diagnosis of diabetic retinopathy.
Materials and Methods: The algorithm can be divided into three main steps. The purpose of the first step or pre-processing is background normalization and contrast enhancement of the images.
The second step aims to detect candidates, i.e., all patterns possibly corresponding to MA, which is achieved using a local radon transform, Then, features are extracted, which are used in the last step to automatically classify the candidates into real MA or other objects using the SVM method. A database of 100 annotated images was used to test the algorithm. The algorithm was compared to manually obtained gradings of these images.
Results: The sensitivity of diagnosis for DR was 100%, with specificity of 90% and the sensitivity of precise MA localization was 97%, at an average number of 5 false positives per image.
Discussion and Conclusion: Sensitivity and specificity of this algorithm make it one of the best methods in this field. Using the local radon transform in this algorithm eliminates the noise sensitivity for MA detection in retinal image analysis.