Figure 3
From: Automatic choroidal segmentation in OCT images using supervised deep learning methods
Example results for the segmentation of a single scan using the RNN 32 × 32, RNN 128 × 32, and RNN 32 × 32 with contrast enhancement (CE). From top to bottom: raw image, combined colour-coded probability maps of the three boundary classes, and boundary delineations where dotted lines are the predicted boundary locations and the solid lines indicate the true boundary locations. Each colour indicates a different boundary; red: ILM, green: RPE and blue: CSI. The effect of increasing the patch size and using contrast enhancement is evident with closer agreement between the true and predicted boundaries.
