Figure 3.

Schematic of the Deep-Learning Model for Automated Segmentation of the Whole Heart Within the Pericardium and Example Automated Segmentation

(A) Block diagram of the Residual U-Net–based convolutional neural network architecture. Details of each layer are provided in the Supplemental Methods. (B) A single CCTA from the ORFAN study demonstrating human expert segmentation as ground truth, automated machine segmentation and a merge. A = anterior; Concat = concatenation; Cov = convolution; I = inferior; L = left; P = posterior; R = right; RelU = rectified linear activation function; S = superior; other abbreviations as in Figures 1 and 2.