Fig. 3
From: Automated diagnosis of plus form and early stages of ROP using deep learning models
(A) In the proposed method for enhanced vessel extraction, a filter is applied twice to the image to improve detail extraction before being processed by the segmentation model. Due to noise in the model’s output, a noise reduction technique is then applied to refine the extracted vessels. (B) The retinal image is directly fed into the segmentation model, where vessel extraction is performed using the standard approach.
