CyclePhase: Robust phase detection in cardiovascular imaging through cyclic motion estimation

Soufiane Ben Haddou, Rudolf Leonardus Mirjam Van Herten, Connie R. Bezzina, Nils Planken, Joost Daemen, Jolanda J. Wentzel, José P. Henriques, Ivana Isgum

Keywords: deep learning, periodic motion estimation, intravascular ultrasound, cine MRI

TL;DR: We propose a unified deep learning framework that models cardiac phase as a continuous cyclic variable, enabling precise and temporally coherent phase detection across IVUS and cine MRI.

Abstract: Accurate cardiac phase detection is essential for cardiovascular imaging applications requiring temporally aligned measurements. While existing methods treat phase detection as discrete frame classification, we propose a fundamentally different approach that models cardiac phase as a continuous cyclic variable on the unit circle. Our method introduces gradient-based input transformations to isolate motion from static anatomy, thereby making it robust to appearance variations, such as calcifications, in intravascular ultrasound (IVUS). Through multi-objective optimization combining temporal consistency via Earth mover's distance with continuous phase regression, we achieve superior performance across both IVUS and cardiac MRI. Experiments demonstrate that explicitly modelling cardiac periodicity yields more accurate and temporally coherent phase detection compared to classification-based approaches, with particular improvements in artefact-heavy clinical scenarios. Our unified framework eliminates the need for modality-specific preprocessing or segmentation masks, providing an end-to-end solution for cardiac motion characterization.

Primary Subject Area: Image Registration

Secondary Subject Area: Unsupervised Learning and Representation Learning

Registration Requirement: Yes

Reproducibility: https://github.com/SoufianeBH/Cyclical-Phase-Detection

Visa & Travel: No

Read CFP & Author Instructions: Yes

Originality Policy: Yes

Single-blind & Not Under Review Elsewhere: Yes

LLM Policy: Yes

Midl Latex Submission Checklist: Ensure no LaTeX errors during compilation., Replace NNN with your OpenReview submission ID., Includes \documentclass{midl}, \jmlryear{2026}, \jmlrworkshop, \jmlrvolume, \editors, and correct \bibliography command., Did not override options of the hyperref package., Did not use the times package., Use the correct spelling and format, avoid Unicode characters, and use LaTeX equivalents instead., Any math in the title and abstract must be enclosed within $...$., Did not override the bibliography style defined in midl.cls and did not use \begin{thebibliography} directly to insert references., Avoid using \scalebox; use \resizebox when needed., Included all necessary figures and removed *unused* files in the zip archive., Removed special formatting, visual annotations, and highlights used during rebuttal., All special characters in the paper and .bib file use LaTeX commands (e.g., \'e for é)., No separate supplementary PDF uploads., Acknowledgements, references, and appendix must start after the main content.

Submission Number: 136