Microaneurysm detection for early diagnosis of diabetic retinopathy

Microaneurysms (MAs) are the first visible sign of diabetic retinopathy (DR), a retinal abnormality which may lead to blindness in diabetes patients. In time and precise MAs detection is very important for early diagnosis of DR and can save patient's vision. In this paper, we present an automated system for accurate and reliable detection of MAs. The proposed system consists of preprocessing, feature extraction and classification stages. The preprocessing step extracts all possible regions which may be considered as MAs from input retinal image and feature extraction stage represents each region with a number of features. A novel hybrid classifier which combines Gaussian mixture model and support vector machine in an ensemble, finally classifies each region as MA or non-MAo The proposed system uses genetic algorithm in order to optimized the weights for hybrid classifier. The evaluation of proposed system is performed using publicly available retinal image database and results are compared with existing techniques to demonstrate the validity of proposed system.