Retinal Fundus Identification Utilizing Supervised and Unsupervised Nature of Deep Neural Network

Eye disease has become serious concern to people, specially who have been suffering from diabetes . Systematic eye diseases are manifest in around the retina of an eye. Digital photography of the retinal images provide a significant information about the disease. This kind of retinal images work as a powerful source for the doctors and ophthalmologist to take meaningful decision about the disease. This kind of Computer Aided Diagnosis(CAD) allow both the doctors and patient to take a second layer of satisfaction. Considering this issues, this paper segmented the retinal vessel images using the state-of-the- art Deep Neural Network(DNN) such as a combination of U-Net and Generative Adversarial Network (GAN) model. As the retinal images suffers due to the Contrast and Illumination problems, this paper also proposed two preprocessing algorithms to reduce this effect.