Fig. 3 | Scientific Data

Fig. 3

From: FIVES: A Fundus Image Dataset for Artificial Intelligence based Vessel Segmentation

Fig. 3

Workflow of the establishment of the proposed dataset. (a) Eight hundred color fundus photographs were collected in the Ophthalmology Center at the Second Affiliated Hospital of Zhejiang University (SHAZU). These photographs comprised 200 images from patients diagnosed with diabetic retinopathy (DR), 200 from patients with glaucoma, 200 from patients with age-related macular degeneration (AMD), and 200 from normal controls. (b) The annotation team was made up of 3 senior annotators and 24 junior annotators who had completed guideline-based training and image annotation tests. (c) The interface of the specifically designed annotating software used in this study. (d) The annotation process consisted of initial annotation by junior annotators and further verification by senior annotators. Each image was randomly assigned to 2 annotators, and the common pixels annotated by both of them were included as the result of initial annotation.

Back to article page