Image-to-Graph Convolutional Network for Deformable Shape Reconstruction from a Single Projection Image

TL;DR

This paper introduces an image-to-graph convolutional network that reconstructs deformable organ shapes from a single X-ray image, demonstrating accurate liver shape recovery with a mean error of 3.6mm.

Contribution

The paper presents a novel IGCN framework that models shape deformation from single-view images using a deformation mapping scheme, advancing deformable shape reconstruction methods.

Findings

Reconstructed liver shapes with a mean error of 3.6mm.

The framework effectively captures respiratory motion of abdominal organs.

The regularized loss function improves reconstruction accuracy.

Abstract

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.