EUVIMEDEUVIMED entdecken
Crossref

Reverse Imaging: Any-Sequence Generalization for Cardiac MRI Segmentation

Yidong Zhao, Yi Zhang, Tongyun Yang, Masa Bozic-Iven, Ayda Arami, Yuchi Han, Orlando Simonetti, Hui Xue, Peter Kellman, Sebastian Weingartner, Qian Tao

IEEE Transactions on Medical Imaging · 2026-11 · 45 (11) · 5366-5379

DOI: 10.1109/tmi.2026.3720503 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Medical Imaging, 2026-11. Verfasst von Yidong Zhao und 10 weiteren Autor:innen. Hauptthemen laut MEDLINE: Heart, Magnetic Resonance Imaging, Image Processing, Computer-Assisted, Cardiac Imaging Techniques. Gefördert von: Delft Artificial Intelligence (AI) Initiative at Delft University of Technology, Dutch Research Council, European Research Council.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Pretrained segmentation models for cardiac magnetic resonance imaging (MRI) often fail to generalize across imaging sequences due to substantial contrast variations. These variations arise from different imaging protocols, yet fundamentally, all contrasts are governed by the same underlying tissue properties, primarily captured by three components: the magnetization strength ( $\mathrm {M}_{\mathrm {{0}}}$ ), $\mathrm {T}_{1}$ , and $\mathrm {T}_{2}$ . Building on this insight, we introduce Reverse Imaging, a physics-driven framework for data augmentation and domain generalization in cardiac MRI. Our method infers tissue properties from observed MR images with annotation by solving an ill-posed nonlinear inverse problem, regularized by a generative prior. The prior is learned from the multiparametric saturation-recovery single shot acquisition (mSASHA) dataset for joint cardiac $\mathrm {T}_{1}$ and $\mathrm {T}_{2}$ mapping. In inference, we characterize imaging sequences as weak, moderate, or strong observations according to the physical information they provide and the degree of ill-posedness. This motivates an iterative prior-learning strategy that uses moderate $\mathrm {T}_{1}$ -mapping observations to alleviate mSASHA data scarcity via pseudo tissue-property estimates. We further integrate MRI physics into posterior inference by expressing the sequence model as a likelihood term guiding the reverse diffusion process. For widely used but weak cine observations, we develop a sequence-specific ControlNet to improve efficiency and spatial consistency. Extensive experiments on eight unseen cardiac MRI sequences with markedly different contrast mechanisms show that Reverse Imaging yields plausible tissue-property estimates, supports synthesis of diverse yet physically consistent contrasts, and improves segmentation robustness under severe cross-sequence shifts.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

HeartMagnetic Resonance ImagingImage Processing, Computer-AssistedCardiac Imaging TechniquesHumansAlgorithms

Förderung

  • Delft Artificial Intelligence (AI) Initiative at Delft University of Technology
  • Dutch Research Council
  • European Research Council
  • Dutch Heart Foundation
  • NVIDIA Academic Grant Program

Zitieren

APA

Zhao, Y., Zhang, Y., Yang, T., Bozic-Iven, M., Arami, A., Han, Y., Simonetti, O., Xue, H., Kellman, P., Weingartner, S., & Tao, Q. (2026). Reverse Imaging: Any-Sequence Generalization for Cardiac MRI Segmentation. IEEE Transactions on Medical Imaging, 45(11), 5366-5379. https://doi.org/10.1109/tmi.2026.3720503

Vancouver

Zhao Y, Zhang Y, Yang T, Bozic-Iven M, Arami A, Han Y, et al. Reverse Imaging: Any-Sequence Generalization for Cardiac MRI Segmentation. IEEE Transactions on Medical Imaging. 2026;45(11):5366-5379. doi:10.1109/tmi.2026.3720503

BibTeX
@article{Zhao2026,
  author = {Zhao, Yidong and Zhang, Yi and Yang, Tongyun and Bozic-Iven, Masa and Arami, Ayda and Han, Yuchi and Simonetti, Orlando and Xue, Hui and Kellman, Peter and Weingartner, Sebastian and Tao, Qian},
  title = {Reverse Imaging: Any-Sequence Generalization for Cardiac MRI Segmentation},
  journal = {IEEE Transactions on Medical Imaging},
  year = {2026},
  volume = {45},
  number = {11},
  pages = {5366--5379},
  doi = {10.1109/tmi.2026.3720503}
}
Kostenlos starten →