2D/3D Reconstruction of Patient-Specific Surface Models and Uncertainty Estimation via Posterior Shape Models

In this paper, we propose a novel method for 3D reconstruction of patient-specific surface models from 2D X-ray images and uncertainty quantification via posterior shape models. Taking the silhouette point cloud generated from biplanar X-ray images as the given partial information, a posterior shape model is constructed to compute the posterior distribution of the surface model given the silhouette. By sampling surface models from the posterior distribution, we can not only compute the patient-specific 3D reconstruction but also quantify the reconstruction uncertainty. Comprehensive experiments were conducted on 25 synthetic cases and 10 cadaveric cases of the proximal femur. Both quantitative and qualitative results demonstrated the effectiveness of the posterior shape model-based reconstruction method.