A novel approach for the automatic detection of microaneurysms in retinal images

Diabetic retinopathy (DR) is the most frequent cause of new cases of blindness among adults aged 20-74 years. Since the presence of microaneurysms (MAs) is usually the first sign of DR, a reliable screening system for the detection of MAs on digital fundus images can provide great assistance to ophthalmologists in difficult diagnoses. This paper presents a novel algorithm to identify the position of MAs on retinal images. The candidates are extracted from a local maxima map, where MAs appear as isolated bright spots. The map is constructed in an angular scanning process of the entire image, recording peaks on one dimensional intensity profiles. Candidates are filtered based on size to eliminate vessel segments. Resulting MAs are selected from the remaining candidates in an additional refined scanning step. The proposed algorithm is well parallelizable and competitive in the field.