Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Unsupervised MRI-US Multimodal Image Registration with Multilevel Correlation Pyramidal Optimization

Jiazheng Wang, Zeyu Liu, Min Liu, Xiang Chen, Hang Zhang

TL;DR

This work targets unsupervised registration between preoperative MRI and intraoperative ultrasound under large tissue deformations and low feature distinctness. It introduces MCPO, a multilevel correlation pyramidal optimization framework that builds modality-independent Mind-SSC features, computes dense correlation volumes, and refines the displacement field via weight-balanced convex optimization across scales, with an optional Adam-based refinement. The method, available in two variants MCPO-rigid and MCPO-deform, achieves top performance in the Learn2Reg 2025 ReMIND2Reg subchallenge (validation and test phases) and demonstrates strong generalization on the Resect dataset with an average TRE of 1.798 mm. The availability of code promotes practical deployment for preoperative-to-intraoperative image registration in surgical navigation and related multimodal tasks.

Abstract

Surgical navigation based on multimodal image registration has played a significant role in providing intraoperative guidance to surgeons by showing the relative position of the target area to critical anatomical structures during surgery. However, due to the differences between multimodal images and intraoperative image deformation caused by tissue displacement and removal during the surgery, effective registration of preoperative and intraoperative multimodal images faces significant challenges. To address the multimodal image registration challenges in Learn2Reg 2025, an unsupervised multimodal medical image registration method based on multilevel correlation pyramidal optimization (MCPO) is designed to solve these problems. First, the features of each modality are extracted based on the modality independent neighborhood descriptor, and the multimodal images is mapped to the feature space. Second, a multilevel pyramidal fusion optimization mechanism is designed to achieve global optimization and local detail complementation of the displacement field through dense correlation analysis and weight-balanced coupled convex optimization for input features at different scales. Our method focuses on the ReMIND2Reg task in Learn2Reg 2025. Based on the results, our method achieved the first place in the validation phase and test phase of ReMIND2Reg. The MCPO is also validated on the Resect dataset, achieving an average TRE of 1.798 mm. This demonstrates the broad applicability of our method in preoperative-to-intraoperative image registration. The code is avaliable at https://github.com/wjiazheng/MCPO.

Unsupervised MRI-US Multimodal Image Registration with Multilevel Correlation Pyramidal Optimization

TL;DR

This work targets unsupervised registration between preoperative MRI and intraoperative ultrasound under large tissue deformations and low feature distinctness. It introduces MCPO, a multilevel correlation pyramidal optimization framework that builds modality-independent Mind-SSC features, computes dense correlation volumes, and refines the displacement field via weight-balanced convex optimization across scales, with an optional Adam-based refinement. The method, available in two variants MCPO-rigid and MCPO-deform, achieves top performance in the Learn2Reg 2025 ReMIND2Reg subchallenge (validation and test phases) and demonstrates strong generalization on the Resect dataset with an average TRE of 1.798 mm. The availability of code promotes practical deployment for preoperative-to-intraoperative image registration in surgical navigation and related multimodal tasks.

Abstract

Surgical navigation based on multimodal image registration has played a significant role in providing intraoperative guidance to surgeons by showing the relative position of the target area to critical anatomical structures during surgery. However, due to the differences between multimodal images and intraoperative image deformation caused by tissue displacement and removal during the surgery, effective registration of preoperative and intraoperative multimodal images faces significant challenges. To address the multimodal image registration challenges in Learn2Reg 2025, an unsupervised multimodal medical image registration method based on multilevel correlation pyramidal optimization (MCPO) is designed to solve these problems. First, the features of each modality are extracted based on the modality independent neighborhood descriptor, and the multimodal images is mapped to the feature space. Second, a multilevel pyramidal fusion optimization mechanism is designed to achieve global optimization and local detail complementation of the displacement field through dense correlation analysis and weight-balanced coupled convex optimization for input features at different scales. Our method focuses on the ReMIND2Reg task in Learn2Reg 2025. Based on the results, our method achieved the first place in the validation phase and test phase of ReMIND2Reg. The MCPO is also validated on the Resect dataset, achieving an average TRE of 1.798 mm. This demonstrates the broad applicability of our method in preoperative-to-intraoperative image registration. The code is avaliable at https://github.com/wjiazheng/MCPO.
Paper Structure (13 sections, 5 figures, 2 tables)

This paper contains 13 sections, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Experiments and Results
  4. Dataset
  5. ReMIND2Reg Sub-challenge.
  6. Resect Dataset.
  7. Implementation Details
  8. Experimental Results
  9. Validation Phase of ReMIND2Reg Sub-challenge.
  10. Resect dataset.
  11. Test Phase of ReMIND2Reg Sub-challenge.
  12. Conclusion
  13. Acknowledgments.

Figures (5)

  • Figure 1: The overall flow of the proposed MCPO method.
  • Figure 2: Visualization results of ReMIND2Reg sub-challenge in Learn2Reg 2025.
  • Figure 3: Visualization results of a large deformation case in the Resect dataset using MCPO-deform.
  • Figure 4: Visualization result of MRI-US fusion for the large deformation case before (left) and after (right) registration using MCPO-deform.
  • Figure 5: Results of methods in the test phase of ReMIND2Reg Sub-challenge. Reported by the organizer.