Synthesizing Retinal and Neuronal Images with Generative Adversarial Nets

A broad range of biomedical images contain thin and long tubular-like foreground objects. They include tubular structured images of various modalities such as magnetic resonance angiography, x-ray angiography, retinal fundus images, as well as cellular neuronal images. Taking retinal fundus images as an example, topological and geometrical properties  (Martinez-Perez et al., 2000) of the vessel structures provide valuable clinical information in diagnosing diseases such as proliferative diabetic retinopathy, glaucoma, and hypertensive retinopathy (Abramoff et al., 2010).
Source: Medical Image Analysis - Category: Radiology Authors: Source Type: research