EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Single-shot multi-slice spiral spatiotemporally encoded MRI with deep-learning-based referenceless distortion correction

Junxian Jin, Wei Wang, Jian Han, Jian Wu, Qinqin Yang, Lin Chen, Shuhui Cai, Hongchang Gao, Congbo Cai, Zhong Chen

Magnetic Resonance Letters · 2026-11 · 200315

DOI: 10.1016/j.mrl.2026.200315 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Magnetic Resonance Letters, 2026-11. Verfasst von Junxian Jin und 9 weiteren Autor:innen. Die Arbeit nennt 72 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Ultrafast magnetic resonance imaging (MRI) is critical for high-temporal-resolution applications, but conventional echo-planar imaging (EPI) is prone to geometric distortion, signal loss and blurring under B 0 inhomogeneity. Spatiotemporally encoded (SPEN) MRI is more robust to field inhomogeneity and chemical-shift artefacts, and non-Cartesian sampling can further improve acquisition efficiency. In this study, we propose a single-shot multi-slice spiral SPEN MRI method, termed MS-sSPEN, together with a referenceless deep-learning-based distortion-correction framework. Through an integrated sequence design combining spiral sampling, segmented multi-slice excitation, and SPEN encoding, MS-sSPEN improves acquisition efficiency, shortens the interval between slice-selective excitations, and reduces radiofrequency energy deposition, thereby alleviating specific absorption rate (SAR) constraints in rapid multi-slice imaging. MRiLab-based physics simulations were used to generate synthetic training data with diverse anatomical textures and random B 0 inhomogeneity fields, enabling geometric distortion correction without additional field-mapping scans or paired real training data. As a preclinical methodological validation, numerical simulations, phantom experiments and in vivo rat brain experiments showed that the proposed method reduced B 0 -inhomogeneity-induced distortion and blurring, simultaneously estimated Δ B 0 maps and corrected each slice within tens of milliseconds after training. The proposed method therefore provides an efficient, robust and referenceless strategy for rapid multi-slice imaging under field-inhomogeneous conditions.

Zitieren

APA

Jin, J., Wang, W., Han, J., Wu, J., Yang, Q., Chen, L., Cai, S., Gao, H., Cai, C., & Chen, Z. (2026). Single-shot multi-slice spiral spatiotemporally encoded MRI with deep-learning-based referenceless distortion correction. Magnetic Resonance Letters, 200315. https://doi.org/10.1016/j.mrl.2026.200315

Vancouver

Jin J, Wang W, Han J, Wu J, Yang Q, Chen L, et al. Single-shot multi-slice spiral spatiotemporally encoded MRI with deep-learning-based referenceless distortion correction. Magnetic Resonance Letters. 2026:200315. doi:10.1016/j.mrl.2026.200315

BibTeX
@article{Jin2026,
  author = {Jin, Junxian and Wang, Wei and Han, Jian and Wu, Jian and Yang, Qinqin and Chen, Lin and Cai, Shuhui and Gao, Hongchang and Cai, Congbo and Chen, Zhong},
  title = {Single-shot multi-slice spiral spatiotemporally encoded MRI with deep-learning-based referenceless distortion correction},
  journal = {Magnetic Resonance Letters},
  year = {2026},
  pages = {200315},
  doi = {10.1016/j.mrl.2026.200315}
}
Kostenlos starten →