Skip to content

Advertisement

Volume 18 Supplement 1

19th Annual SCMR Scientific Sessions

  • Oral presentation
  • Open Access

Motion compensated reconstruction from free breathing 2D radial cardiac MRI data

  • 4, 1,
  • 4,
  • 4, 5,
  • 2,
  • 3 and
  • 1
Journal of Cardiovascular Magnetic Resonance201618 (Suppl 1) :O110

https://doi.org/10.1186/1532-429X-18-S1-O110

  • Published:

Keywords

  • Cardiac Magnetic Resonance
  • Reconstruction Strategy
  • Respiratory Motion Artifact
  • Cardiac Magnetic Resonance Data
  • Golden Angle

Background

Cardiac magnetic resonance imaging (CMR) is affected by both cardiac and respiratory motion. While ECG-gated imaging within a breath hold is often the method of choice to limit motion-related artifacts, free-breathing methods are favorable in patients with limited breath hold capability. Free-breathing approaches require either rapid single-shot scans to reduce respiratory motion artifacts at the expense of spatial resolution or higher resolution segmented respiratory-gated scans at the expense of scan time efficiency. Previous work has exploited the favorable properties (e.g., motion robustness, uniform sampling density) of Golden Angle [1] radial sampling including motion. Recently introduced motion compensated reconstructions [2,3] have been applied to various clinical applications. In this work, we propose to combine a 2D radial Golden Angle data acquisition scheme with a recently developed motion compensated reconstruction strategy [4] to obtain high-resolution motion compensated CMR data from time-efficient cardiac-gated free-breathing exams.

Methods

A cardiac gated 2D golden angle radial spoiled gradient echo sequence with the following parameters was used: α = 15°, BW = ± 125 kHz, 256 readout points, TR = 4.26 ms, TE = 1.50 ms, FOV = 360 × 360 mm2. Data were acquired in diastole, total scan time was 50s. The respiratory belt signal was recorded synchronously with the MR acquisition. Three images were reconstructed from free-breathing data corresponding to 50s, 6s, and to 3s using a non-Cartesian iterative SENSE reconstruction [5]: 1) Combining all free-breathing data without motion management (FB), 2) Retrospective gating (respiratory belt signal) using data closest to end-expiration (RG), 3) Motion Compensated Reconstruction (MCR). The MCR is obtained in 4 steps: a) clustering of the all data into 6 respiratory bins according to the respiratory belt signal, b) independent reconstruction of the 6 bins, c) extraction of the motion between the bins by applying a non-rigid registration, c) model based reconstruction utilizing a motion-compensated SENSE-like reconstruction [6,7] (see also Figure 1).
Figure 1
Figure 1

Scheme of the motion compensated reconstruction (MCR).

Results

Figure 2 compares reconstruction strategies. Combining all FB data into one dataset leads to significant blurring. RG improves apparent sharpness; however, small anatomical structures are best visualized in the MCR. The benefit of the MCR is more pronounced in the 6s and 3s examples.
Figure 2
Figure 2

Comparison of reconstruction results.

Conclusions

Motion compensated reconstruction is a promising technique to remove respiratory motion. By combining multiple respiratory bins into one spatially high-resolved, motion-corrected dataset, images with high visual sharpness are obtained in a time-efficient manner. Cardiac clinical applications such as function, perfusion, or late enhancement can potentially benefit from the presented data acquisition and reconstruction strategy and are are currently under investigation.

Authors’ Affiliations

(1)
Cardiac Center of Excellence, GE Healthcare, Garching, Germany
(2)
Medical Physics, University of Wisconsin, Madison, WI, USA
(3)
Radiology, University of Wisconsin, Madison, WI, USA
(4)
GE Global Research, Garching, Germany
(5)
Computer Science, Technische Universität München, Munich, Germany

Copyright

Advertisement