Deep-Learning-Based Choroid Vessel Segmentation Using Feature Maps of ResUNet GAN Autoencoder Architecture

We propose a deep-learning approach for choroid vessel segmentation in retinal OCT, leveraging the intermediate feature maps of a ResUNet-based GAN autoencoder to extract complex vascular structures. The method underpins automated estimation of the Choroidal Vasculature Index (CVI) and supports clinical analysis of choroidal vasculature.

In Investigative Ophthalmology & Visual Science, vol. 66, no. 8, pp. 587–587, 2025. The Association for Research in Vision and Ophthalmology.