RRN:Recursive Refinement Network: Learning based deformable image registration

By Xinzi He, Jia Guo, Xuzhe Zhang, Hanwen Bi, Sarah Gerard, David Kaczka, Amin Motahari, Eric Hoffman, Joseph Reinhardt, R. Graham Barr, Elsa Angelini, and Andrew Laine.

Paper link: [arXiv] https://arxiv.org/pdf/2106.07608.

Unsupervised learning-based medical image registration approaches have witnessed rapid development in recent years. We propose to revisit a commonly ignored while simple and well-established principle: recursive refinement of deformation vector fields across scales. We introduce a recursive refinement network (RRN) for unsupervised medical image registration, to extract multi-scale features, construct normalized local cost correlation volume and recursively refine volumetric deformation vector fields. RRN achieves state of the art performance for 3D registration of expiratory-inspiratory pairs of CT lung scans. On DirLab COPDGene dataset, RRN returns an average Target Registration Error (TRE) of 0.83 mm, which corresponds to a 13% error reduction from the best result presented in the leaderboard 4. In addition to comparison with conventional methods, RRN leads to 89% error reduction compared to deep-learning-based peer approaches.

Requirements

The code has been tested with PyTorch 1.6 and Cuda 10.1.

conda create --name rrn
conda activate rrn
conda install pytorch=1.6.0 torchvision=0.7.0 cudatoolkit=10.1 matplotlib tensorboard scipy opencv nibabel -c pytorch

Required Data

To evaluate/train RRN, you will need to download the required datasets.

• DirLab COPDGene

Reference

Thanks to previous open-sourced repo:
VoxelMorph
UFlow