The automated detection of proliferative diabetic retinopathy using dual ensemble classification

Objective: Diabetic retinopathy (DR) is a retinal vascular disease that is caused by complications of diabetes. Proliferative diabetic retinopathy (PDR) is the advanced stage of the disease which carries a high risk of severe visual impairment. This stage is characterized by the growth of abnormal new vessels. We aim to develop a method for the automated detection of new vessels from retinal images. Methods: This method is based on a dual classification approach. Two vessel segmentation approaches are applied to create two separate binary vessel maps which each hold vital information. Local morphology, gradient and intensity features are measured using each binary vessel map to produce two separate 21-D feature vectors. Independent classification is performed for each feature vector using an ensemble system of bagged decision trees. These two independent outcomes are then combined to a produce a final decision. Results: Sensitivity and specificity results using a dataset of 60 images are 1.0000 and 0.9500 on a per image basis. Conclusions: The described automated system is capable of detecting the presence of new vessels.