Image-to-Graph Convolutional Network for Deformable Shape Reconstruction from a Single Projection Image
This addresses shape reconstruction for image-guided medical interventions, but appears incremental as it builds on existing deep learning and graph convolutional methods.
The paper tackles the problem of reconstructing deformable organ shapes from a single 2D X-ray image, achieving a mean distance error of 3.6mm for liver shapes in experiments on respiratory motion.
Shape reconstruction of deformable organs from two-dimensional X-ray images is a key technology for image-guided intervention. In this paper, we propose an image-to-graph convolutional network (IGCN) for deformable shape reconstruction from a single-viewpoint projection image. The IGCN learns relationship between shape/deformation variability and the deep image features based on a deformation mapping scheme. In experiments targeted to the respiratory motion of abdominal organs, we confirmed the proposed framework with a regularized loss function can reconstruct liver shapes from a single digitally reconstructed radiograph with a mean distance error of 3.6mm.