On Classifying Diabetic Patients’ with Proliferative Retinopathies via a Radial Basis Probabilistic Neural Network

Diabetic retinopathies have to be detected early and treated to avoid serious damages to patients’ retina. A severe progress of diabetes can deteriorate human vision and the effects of a Proliferative Diabetic Retinopathy (PDR) could appear in fundus images, showing a neovascularization that can rise abruptly. Until now only some network models for classifying presence/absence of PDR have been faced by means of PNNs or SVMs. In this paper a first approach to follow diabetic patients affected by early PDR via a novel neural classifier based on a Fundus Image Preprocessing Subsystem and a Radial Basis Probabilistic Neural Network (RBPNN) is presented. The proposed classifier aims at classifying a certain number of diabetic patients by means of their accurately preprocessed digital fundus images and could support their follow-up paths in alerting if variations in retinal vasculature of classified PDR should occur.