Deep Belief Network based Diabetic Maculopathy Detection and Classification using Modified Chicken Swarm Algorithm

One of the most frequent causes of visual impairment in people with diabetes mellitus is diabetic maculopathy (DM). The macula is the important a part of the retina accountable for sharp vision. Diabetic maculopathy is a type of diabetic eye disease that affects the macula. The most prevalent and typical sign of diabetic maculopathy is diabetic macular edema(DME). In addition, people with diabetics one of the common cause for DME is visual impairment. Therefore, it is of the utmost important to detect DME region early and accurately. On the other hand, Ophthalmologists widely use optical coherence tomography (OCT) as an imaging tool to identify retinal diseases. In this paper, Chicken Swarm Optimization driven Deep Belief Network is developed for DME classification. In this method, OCT images are taken into account for the effective classification of the DME process. In this classification process, the layer segmentation stage involves segmenting 12 layers and 13 borders. Moreover, the Deep Belief Network classifier is applied in order to classify the images of OCT as normal one or DME affected region. The GAN classifier is trained by Chicken swarm algorithm such that the efficiency of classification is enhanced. This developed approach outperforms with other existing DME approach with respect to accuracy, specificity and sensitivity of 94.24%, 94% and 94% respectively.