Figure 1.

Proposed deep learning method. A, Schematic view of the deep learning model. The numbers written beside each layer represent the number of feature maps × width (pixels) × height (pixels). B, Detailed expositions of the DownBlock, UpBlock, multiple dilated convolutional (MDC) block, and signs (arrows and layers). C, Representative input image, manually annotated ground truth, and automatic vessel segmentation of the digital retinal images for vessel extraction dataset (top row) and representative input image, manually annotated ground truth, and automatic vessel segmentation of the Hokudai dataset (bottom row).