Figure 1:
The schematic diagram of the proposed Joint craniomaxillofacial bone Segmentation and landmark Digitization (JSD) framework. There are two major components, including (1) estimation of displacements from voxels to landmarks via the first-stage fully convolutional network (FCN1), and (2) joint bone segmentation and landmark digitization via the second-stage FCN (FCN2). The locations of 15 anatomical landmarks are also illustrated in this figure. FCN: fully convolutional neural network.