Deep Active Latent Surfaces for Medical Geometries

Abstract

Shape priors have long been known to be effective when reconstructing 3D shapes from noisy or incomplete data. When using a deep-learning based shape representation, this often involves learning a latent representation, which can be either in the form of a single global vector or of multiple local ones. The latter allows more flexibility but is prone to overfitting. In this paper, we advocate a hybrid approach representing shapes in terms of 3D meshes with a separate latent vector at each vertex. During training the latent vectors are constrained to have the same value, which avoids overfitting. For inference, the latent vectors are updated independently while imposing spatial regularization constraints. We show that this gives us both flexibility and generalization capabilities, which we demonstrate on several medical image processing tasks. * Equal supervision.Preprint. Under review.

ShapeMed-Knee: A Dataset and Neural Shape Model Benchmark for Modeling 3D Femurs

“…These data demonstrate that the NSM faithfully reconstructs plausible anatomical surfaces while filling in the corrupted regions with reconstructions based on the learned priors. This finding supports previous research which indicates that NSMs can be used as a means of refining automated segmentations to ensure anatomic plausibility [38].…”

“…During training, a single latent vector was used for all points, while during inference, latents vary over the surface to increase expressivity. They showed better reconstruction than DeepSDF and improved segmentation results [38]. Ludke et al used a neural flow deformer to fit a NSM by deforming coordinates from a template shape to the target, outperforming a conventional SSM in terms of surface reconstruction and simple OA classification [39].…”