Area Agnostic Pipeline for Retina Vessel Segmentation
Authors: Benjamin Hou
Summary: Automated segmentation of retina vessels performs a pivotal function in scientific analysis of prevalent eye illnesses, resembling, Diabetic Retinopathy or Age-related Macular Degeneration. As a result of advanced development of blood vessels, with drastically various thicknesses, correct vessel segmentation may be fairly a difficult activity. On this work we present that it’s potential to realize close to state-of-the-art efficiency, by crafting a cautious thought pre-processing pipeline, with out having to resort to advanced networks and/or coaching routines. We additionally present that our mannequin is ready to preserve the identical excessive segmentation efficiency throughout completely different datasets, very poor high quality fundus pictures, in addition to pictures of extreme pathological instances. Code and fashions featured on this paper may be downloaded from http://github.com/farrell236/retina_segmentation. We additionally reveal the potential of our mannequin at http://lazarus.ddns.net:8502.
