Landmark-Free Preoperative-to-Intraoperative Registration in Laparoscopic Liver Resection

TL;DR

This paper tackles the challenge of registering preoperative liver models to intraoperative views without relying on landmarks. It introduces Self-P2IR, a landmark-free, self-supervised framework that redefines preoperative-to-intraoperative fusion as a 3D-3D registration problem, decomposed into rigid and non-rigid stages. The rigid stage uses a feature-disentangled transformer for robust correspondences, while the non-rigid stage employs a Structure-Regularized Shape Adaptation with a low-rank transformer to guide plausible deformation, all validated with an in-vivo dataset P2I-LReg and synthetic data. Results show superior performance over state-of-the-art methods in both 3D-3D and 3D-2D registrations, with strong robustness to noise and occlusion, and positive surgeon feedback, underscoring potential clinical impact.

Abstract

Liver registration by overlaying preoperative 3D models onto intraoperative 2D frames can assist surgeons in perceiving the spatial anatomy of the liver clearly for a higher surgical success rate. Existing registration methods rely heavily on anatomical landmark-based workflows, which encounter two major limitations: 1) ambiguous landmark definitions fail to provide efficient markers for registration; 2) insufficient integration of intraoperative liver visual information in shape deformation modeling. To address these challenges, in this paper, we propose a landmark-free preoperative-to-intraoperative registration framework utilizing effective self-supervised learning, termed \ourmodel. This framework transforms the conventional 3D-2D workflow into a 3D-3D registration pipeline, which is then decoupled into rigid and non-rigid registration subtasks. \ourmodel~first introduces a feature-disentangled transformer to learn robust correspondences for recovering rigid transformations. Further, a structure-regularized deformation network is designed to adjust the preoperative model to align with the intraoperative liver surface. This network captures structural correlations through geometry similarity modeling in a low-rank transformer network. To facilitate the validation of the registration performance, we also construct an in-vivo registration dataset containing liver resection videos of 21 patients, called \emph{P2I-LReg}, which contains 346 keyframes that provide a global view of the liver together with liver mask annotations and calibrated camera intrinsic parameters. Extensive experiments and user studies on both synthetic and in-vivo datasets demonstrate the superiority and potential clinical applicability of our method.

Figures (16)

• Figure 1: Illustration of (a) challenges of landmarks in laparoscopic scenarios; (b) ambiguity of landmark annotations, with 2D landmarks varying by camera view; (c) our registration method without landmarks or interactive inputs for inference.
• Figure 2: Illustration of our P2I-LReg dataset.
• Figure 3: Overview of the proposed Self-P2IR. The synthetic data is firstly used to train the rigid registration network with a Feature Disentangled Transformer. For non-rigid registration, Structure-Regularized Shape Adaptation (SRSA) is designed to estimate the deformation field, comprising a deformation decomposition pyramid and a low-rank structure similarity learning network. During self-training on in-vivo data, a differentiable renderer is adopted to create the deformed, pose-dependent 2D liver mask, and 3D point cloud to compute the loss between rendered outputs and masks as well as the reprojected point cloud.
• Figure 4: Illustration of our feature disentangled transformer structure.
• Figure 5: The low-rank structure similarity learning network.