DreamReg: Belief-Driven World Model for 2D-3D Ultrasound Registration

2026-06-17

AI Summary

This work addresses the challenges of partial observability, speckle noise, and probe-motion dependency in intraoperative 2D ultrasound to preoperative 3D volume registration by proposing a belief-based world model framework. The approach formulates registration as a sequential belief-updating process over rigid transformations, leveraging an implicit belief state that fuses historical observations and pose information. Coupled with a learned dynamics model, it employs an internal imagination mechanism during inference to iteratively predict probe motion and corresponding observations, yielding adaptive and temporally consistent alignment. Innovatively incorporating clinical scanning behavior simulation during training and an observation-conditioned pose refinement strategy, the method achieves state-of-the-art performance on the CAMUS and u-RegPro datasets, demonstrating both real-time capability and robustness suitable for surgical navigation.

Abstract

Ultrasound (US) is widely used for surgical navigation, yet real-time registration between intraoperative 2D slices and preoperative 3D volumes remains challenging due to partial observability, speckle noise, and the action-dependent US acquisition. Existing methods are one-shot or short-horizon, making it hard for them to gather evidence over time or capture how surgeons adjust probe motion based on on-screen feedback. We propose DreamReg, a belief-driven world-model framework that formulates 2D-3D registration as belief updating over rigid transformations. DreamReg maintains a latent belief state that summarizes past observations and poses information, and continuously refines the transformation through learned dynamics as new slices arrive. During training, DreamReg is exposed to probe-motion trajectories that mimic clinical scanning behavior and learns to update its belief by conditioning pose refinement on the current US observation. During inference, DreamReg refines registration via internal imagination: it rolls out the learned world model to simulate candidate probe motions and their predicted observations, and integrates these imagined outcomes to converge to an accurate rigid transformation. Experiments on CAMUS and u-RegPro datasets demonstrate improved robustness and competitive registration accuracy for real-time guidance compared with state-of-the-art methods.

Problem

• 2D-3D ultrasound registration
• real-time registration
• partial observability
• speckle noise
• action-dependent acquisition

Innovation

• belief-driven world model
• 2D-3D ultrasound registration
• internal imagination
• rigid transformation refinement
• sequential pose estimation