Unsupervised motion correction for real-time free-breathing flow acquisitions: high SNR, single heart-beat, pseudo-breathheld flow images easy to quantify
Haris Saybasili1 and Ning Jin2

1MR R&D, Siemens Healthcare USA, Inc., Chicago, IL, United States, 2MR R&D, Siemens Healthcare USA, Inc., Columbus, OH, United States

Synopsis

Flow quantification on uncooperative patients that cannot hold their breath may be performed by real-time free-breathing flow acquisitions that generally cover multiple heart-beats. However, heart-rate fluctuations, low SNR, and respiratory motion associated with such acquisitions may impede flow quantification process. In this study, we are extending our unsupervised motion correction method for real-time free-breathing cine imaging to flow imaging, and comparing the results with gold standard ECG gated, breath-held, segmented flow acquisition. Preliminary results indicate net forward volumes in agreement with the reference, with comparable image quality.

TARGET AUDIENCE

Cardiovascular MRI, Radiologists, Cardiologists

PURPOSE

ECG gated, breath-held segmented acquisitions are reference standard for cardiac flow quantification with good spatial/temporal resolutions. However, such techniques do not perform well with uncooperative patients that cannot hold their breath, or patients with arrhythmia[NJ1] . Real-time free-breathing acquisitions are viable alternatives, but with the expense of lower SNR and decreased spatial/temporal resolution. Additionally, real-time acquisitions usually covers multiple heartbeats, which makes it difficult to quantify stroke volume or regurgitation fraction. We previously proposed an unsupervised motion correction technique that reconstructed single heart-beat, pseudo-breathheld cardiac cine images from real-time free-breathing acquisitions that covered multiple heart-beats without human interaction (1). In this work, we extend this work and apply it in free-breathing real-time flow acquisitions (2) to generate pseudo-breathhold magnitude and phase flow images in a single heartbeat. The proposed approach was evaluated in healthy volunteers to measure flow in the ascending aorta (AO) and main pulmanory artery (MPA) Net forward flow volume measured from real-time acquisition with motion correction reconstruction was compared to the reference ECG-gated segmented acquisition during breathhold. [NJ1]How do you think this technique would work with patients with arrhythmia? I am just saying segmented is not good for arrhythmias J Real-time works fine, and moco will work as well. Next step for me is to integrate arrhythmia handling.

METHODS

Segmented gradient echo (GRE) and free-breathing real-time echo-planar-imaging (EPI) flow images were acquired in four healthy volunteers (with IRB approval) using a clinical 1.5T MR scanner (MAGNETOM Aera, Siemens Healthcare, Erlangen, Germany). Images were acquired in AO and MPA. Segmented flow images were acquired with 4 lines/segment, 192x120 imaging matrix, BW = 449 Hz/pixel, resulting a temporal resolution of 37 ms and a total acquisition time of 18 heartbeats TGRAPPA factor 3 with shared velocity encoding (2) was used during real-time data acquisitions with EPI factor = 7, 128 * 84 imaging matrix, BW = 2000 Hz/pixel, resulting in a temporal resolution of 32 ms. The real-time images were acquired covering 16 to 20 beats per slice, and the data were processed in an unsupervised fashion using our method as described in (1). Both conventional segmented and motion corrected real-time single heartbeat flow images were analyzed to evaluate net forward volume (syngo Argus, Siemens Healthcare, Erlangen, Germany).

RESULTS

AO and MPA images reconstructed from reference standard, and real-time free-breathing data sets with and without motion correction are presented in Figure 1. Motion corrected images were superior to the original real-time images in terms of SNR. Although low in spatial resolution compared to the reference standard, quality of the motion corrected real-time images was comparable to segmented breath-hold images. The net forward flow volume from the segmented BH and real-time moco approaches was also compared using pair-wise t-tests. There was no significant difference between the values measured by two approaches (p = 0.979). Results are depicted in Fig. 2. A basic background phase correction was applied to motion-corrected images by selecting a reference region-of-interest in the nearby static tissue. Please note that our post-processing tool was applied to reconstructed DICOM images. Therefore, possible aliasing due to inadequate velocity setting might negatively affect the results. This work will be extended to work on original complex images to improve the overall robustness of the aforementioned automated post-processing technique.

CONCLUSION

We demonstrate the feasibility to apply our unsupervised motion correction technique to real-time free-breathing flow acquisitions. This technique generates pseudo breath-held images in a single-heart beat with high-SNR and makes the workflow for flow quantification quite easy. Compared to the referenced standard, the image quality was comparable and net flow volume was in good agreement.

Acknowledgements

No acknowledgement found.

References

(1) Saybasili, et al. ISMRM 2015, #1230.

(2) Lin, et al. Magn Reson Med 2012 Sep; 68(3):703-10.

Figures

Fig.1. Flow images (MPA, AO) reconstructed from ECG gated breath-hold (BH) segmented and free-breathing (FB) real-time acquisitions with and without motion correction (MOCO).

Fig. 2. Comparison of flow values calculated from motion-corrected free-breathing real-time (RT) flow images, and segmented breath-held (BH) ECG gated images.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
3212