Automatic Classification of Macular Edema in Retinal Fundus Images based on Deep Belief Networks

This paper presents a novel Deep Belief Network (DBN) for the automatic classification of macular edema in retinal fundus images. The method involves a preprocessing step that applies a vesselness measure for vessel detection and also to be used as input features for the DBN model. The database of retinal fundus images consists of 200 RGB images (100 with macular edema, and the remaining with healthy patients), which were manually classified by a specialist in Ophthalmology. In order to evaluate the performance of the proposed method, the accuracy metric has been adopted. In addition, the proposed DBN is compared with a set of state-of-the-art convolutional neural networks. According to the experimental results, the proposed method achieved suitable performance in terms of classification accuracy and computational time with 0.975 and 202.7 seconds, respectively.