Preprint
Recursive Refinement Network for Deformable Lung Registration between Exhale and Inhale CT Scans
ArXiv.org
06/14/2021
DOI: 10.48550/arxiv.2106.07608
Abstract
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. In addition to comparison
with conventional methods, RRN leads to 89% error reduction compared to
deep-learning-based peer approaches.
Details
- Title: Subtitle
- Recursive Refinement Network for Deformable Lung Registration between Exhale and Inhale CT Scans
- Creators
- Xinzi HeJia GuoXuzhe ZhangHanwen BiSarah GerardDavid KaczkaAmin MotahariEric HoffmanJoseph ReinhardtR. Graham BarrElsa AngeliniAndrew Laine
- Resource Type
- Preprint
- Publication Details
- ArXiv.org
- DOI
- 10.48550/arxiv.2106.07608
- ISSN
- 2331-8422
- Language
- English
- Date posted
- 06/14/2021
- Academic Unit
- Roy J. Carver Department of Biomedical Engineering; Radiology; Anesthesia; Internal Medicine
- Record Identifier
- 9984320090702771
Metrics
193 Record Views