A Feature Selection Strategy to Optimize Retinal Vasculature Segmentation

Diabetic retinopathy (DR) is a complication of diabetes mellitus that appears in the retina. Clinitians use retina images to detect DR pathological signs related occlusion tiny blood vessels. Such brings degenerative cycle between breaking off and new generation thinner weaker This research aims develop suitable retinal vasculature segmentation method for improving screening procedures by means computer-aided diagnosis systems. The vessel methodology relies on an effective feature selection based Sequential Forward Selection, using error rate decision tree classifier evaluation function. Subsequently, classification process performed three alternative approaches: artificial neural networks, trees support vector machines. proposed validated publicly accessible datasets private one provided Hospital Sant Joan Reus. In all cases we obtain average accuracy above 96% with sensitivity 72% process. Compared state-of-the-art, our approach achieves same performance as other methods need more computational power. Our significantly reduces number features used from 20 5 dimensions. implementation classifiers confirmed five selected have good effectiveness, independently algorithm.