Macular edema computer-aided evaluation in ocular vein occlusions

This paper is concerned with the use of digital fundus imaging to detect, quantify, and follow up macular angiographic leakage due to retinal vein occlusions. Images were matched automatically. We detected those pixels with a high increment in gray level within the closest area to the foveal center. Binary images displaying leakage were obtained. The procedure was checked against two observers' agreement. Twenty-one angiographic studies were collected. Two images of each sequence were selected for digitalization. Numerical descriptors of the leakage were proposed and quantification plots were designed for each pair of images. Interobserver concordance ranged between 82 and 98% when manually detected leakage was compared with computer segmented areas. The detection and quantification of leakage areas may serve as a guide for severity evaluation and treatment planning. Moreover, they permit a precise follow-up of macular edema.