2007 Presentations (Communicative Events)

Retinal vessel segmentation using multi-scale wavelet frame analysis

Lee, Noah; Laine, Andrew F.; Smith, R. Theodore; Busuoic, Mihai

Fundus imaging is a non-invasive technique for in vivo ophthalmoscopic inspection of retinal disorders. Quantitative information about the vascular network can facilitate clinical diagnosis of retinal diseases [1]. Goal: Segmentation of the vascular network in fundus images for further quantification and post processing as a binary classification into object and background. Approach: We perform an over-complete multi-scale wavelet frame expansion with selective channel rejection in the decomposition tree. Remaining channels undergo wavelet shrinkage and enhancement to separate retinal objects from background. Results: Comparison to expert gradings on a pixel by pixel basis show mean sensitivity and specificity of 0.8 and 0.9.