CineMorph: Learning Time-Continuous Motion Field for Motion Tracking on Cine Magnetic Resonance Images

Dongsheng Ruan, Zheng Yuan, Xiaowei Ruan, zhangxiaolin, Chengjin Yu, Mingfeng Jiang, Huafeng Liu

Keywords: Motion tracking, cine MRI, unsupervised learning, diffeomorphic

Abstract: Tracking cardiac motion using cine magnetic resonance imaging (cine MRI) is essential for evaluating cardiac function and diagnosing cardiovascular diseases. Current methods for cardiac motion tracking depend on scaling and squaring (SS) integration to learn discrete Lagrangian motion fields. However, this reliance hinders the effective exploitation of temporal continuity, leading to inadequate tracking accuracy. In this paper, we introduce a novel unsupervised learning method, CineMorph, to achieve temporally continuous cardiac motion tracking in cine MRI image sequences. Our approach integrates a frame-aware UNet with a series of time-continuous Transformer blocks to learn temporally continuous intra-frame motion fields, which are then assembled into time-continuous Lagrangian motion fields. To ensure the diffeomorphism property, we implement semigroup regularization to constrain our model, thus eliminating the reliance on SS integration. We evaluate our method on the public Automatic Cardiac Diagnostic Challenge (ACDC) dataset. The experimental results show that our method outperforms the existing state-of-the-art methods and achieves state-of-the-art performance with a mean DICE score of $83.6\%$ and a mean Hausdorff distance of $3.4$ mm.

Primary Area: applications to computer vision, audio, language, and other modalities

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 8733