Image Magnification Network for Vessel Segmentation in OCTA Images

Optical coherence tomography angiography (OCTA) is a novel non-invasive imaging modality that allows micron-level resolution to visualize the retinal microvasculature. The vessel segmentation in OCTA images still an open problem, and especially thin dense structure of capillary plexus important challenge this problem. In work, we propose image magnification network (IMN) for images. Contrary U-Net with down-sampling encoder up-sampling decoder, proposed IMN adopts design encoding then decoding. This capture more low-level details reduce omission small structures. experimental results on three datasets show average dice score 90.2% achieves best performance Besides, also demonstrate superior cross-field skeleton extraction.