Exudate segmentation in fundus images using an ant colony optimization approach

The leading cause of new blindness and vision defects in working-age people, diabetic retinopathy is a serious public health problem in developed countries. Automatic identification of diabetic retinopathy lesions, such as exudates, in fundus images can contribute to early diagnosis. Currently, many studies in the literature have reported on segmenting exudates, but none of the methods performs as needed. Moreover, several approaches were tested in independent databases, and the approach’s capacity to generalize was not proved. The present study aims to segment exudates with a new unsupervised approach based on the ant colony optimization algorithm. The algorithm’s performance was evaluated with a dataset available online, and the experimental results showed that this algorithm performs better than the traditional Kirsch filter in detecting exudates.