Retinal Vasculature Segmentation Using Wavelets and Supervised Classification

Segmentation of the retinal vasculature is a first step towards automated screening for diabetic retinopathy, a leading cause of adult blindness, but which can be prevented if identified early enough. This paper summarizes the retinal vasculature segmentation approach developed combining the twodimensional continuous wavelet transform and supervised pixel classification. The open source software developed for testing and demonstration is also described. Experimental evaluation using receiver operating characteristic (ROC) analysis on two public image databases yields areas under the ROC curves slightly superior to those presented by other state-of-the-art methods, while minimizing the need for user interaction and showing efficient computing times.