Fig. 2: Results of the segmentation network. | Nature Medicine

Fig. 2: Results of the segmentation network.

From: Clinically applicable deep learning for diagnosis and referral in retinal disease

Fig. 2

Three selected two-dimensional slices from the n = 224 OCT scans in the segmentation test set (left) with manual segmentation (middle) and automated segmentation (right; detailed color legend in Supplementary Table 2). a, A patient with diabetic macular edema. b, A patient with choroidal neovascularization resulting from age-related macular degeneration (AMD), demonstrating extensive fibrovascular pigment epithelium detachment and associated subretinal fluid. c, A patient with neovascular AMD with extensive subretinal hyperreflective material. Further examples of the variation of pathology with model segmentation and diagnostic performance can be found in Supplementary Videos 19. In all examples the classification network predicted the correct diagnosis. Scale bars, 0.5 mm.

Back to article page