A quantification framework for post-lesion neo-vascularization in retinal angiography

Abstract

We describe an image processing framework designed to detect and quantify the genesis of microscopic choroidal blood vessels. We used fluorescein angiography to monitor the dynamic of neo-vascularization of the retina after inducing lesions with a calibrated laser pulse. The angiogenesis can be revealed by an increase in the overall fluorescence level and/or diffusion size of the lesion. The proposed framework allows measuring both features from mis-aligned angiograms acquired with different gains and contrasts. It consists in aligning all the images, homogenizing their intensity characteristics and segmenting the lesions. In particular, we implemented a level set segmentation algorithm to delineate the diffusion area. We show that our framework allows detecting neo-vascularization when one of these features changes by less than 10%.