Introduction

Conventional breath-hold single phase imaging in gadoxetic acid-enhanced liver MR is known to produce artifact in hepatic arterial phase images due to transient dyspnea. Triple arterial phase imaging can acquire an arterial phase image unaffected by severe motion, but data acquisition results in lower SNR due to reduced temporal window¹. Free-breathing and continuous acquisition using golden-angle radial sparse parallel (GRASP) is an alternative and has the potential to reduce motion artifact in arterial phase imaging². Accurate timing for capturing arterial phase is important in reconstruction of continuously acquired data. In this abstract we study a retrospective reconstruction technique that first identifies the timing of contrast enhancement, adjusts time offset in raw data, and reconstructs arterial phase images with different temporal windows.

Methods

Gadoxetic acid-enhanced liver MRI data were acquired during free breathing using a 3D stack-of-stars golden-angle radial sequence on a Philips Achieva 3T scanner. Sequence parameters were as follows: scan time = 1:04, TR/TE = 3.5/1.51 ms, total radial spokes = 379, partial k-space along slice encode = 41/52, flip angle = 10°, image matrix = 504 x 504 x 52, reconstruction voxel size = 0.73 x 0.73 x 2.0 mm. We scanned five healthy volunteers, and the MRI protocol included other routine abdominal MR sequences: T2 weighted imaging, diffusion weighted imaging, eTHRIVE DCE imaging, etc. The BART parallel imaging compressed sensing reconstruction was adopted for the improvement of spatial and temporal resolution and was implemented in MATLAB³. (Identification of maximal aortic enhancement frame) We reconstructed a single slice with temporal resolution of 21 spokes (i.e., 3.5 sec) and measured average signal in the aorta region of interest (ROI) at every frame. The time frame corresponding to the maximum aorta signal served as a reference for determination of arterial phase images. (Retrospective reconstruction with different temporal windows) Based on the peak enhancement frame, we adjusted time offset in the raw data and reconstructed images using parallel imaging compressed sensing. Temporal windows of 55 and 70 contiguous spokes were considered for image quality comparison. Temporal total variation was used for reconstruction.

Results and Discussion

The aorta ROI is shown in Fig 1a. The time intensity curve shows the pattern of contrast enhancement followed by contrast decay (Fig 1b). Although the 21-spokes reconstruction is highly under-sampled, it produces acceptable image quality in the aorta ROI and enables the identification of peak arterial enhancement with high temporal resolution. The peak enhancement frame served as a reference time point. Based on the reference time point (see the red dashed line in Fig 1c), we assigned the temporal frames in the raw data. Fig 2 shows comparison of different temporal resolution reconstructions from the same continuously acquired raw data. The 55-spokes reconstruction shows that the enhancement of hepatic vessel first occurs in 10.1 – 20.2 seconds while the 70-spokes reconstruction does exhibit the enhancement pattern in 0-12.8 seconds. This indicates that the enhancement in the hepatic vessel starts to occur between 10.1 and 12.8 seconds. The 70-spokes reconstruction suggests that higher SNR image can be obtained with longer temporal window in late hepatic arterial phase in which its contrast change is relatively slower than early arterial phase. In this work, we have not addressed artifact issue resulting from respiratory motion. We will incorporate a self-gating technique⁴ into reconstruction and study its effect on image quality.

Conclusion

We have demonstrated that the identification of peak aortic enhancement is feasible with high temporal resolution reconstruction of continuously acquired golden angle radial sequence data. The peak enhancement time information can provide a reference time point. Subsequent retrospective reconstructions with different temporal windows offer flexibility in selecting images from early arterial and late arterial phases.

Acknowledgements

No acknowledgement found.

References

[1] Pietryga et al., Radiology 2014:271(2):426-434. [2] Chandarana et al., Investigative Radiology 2013:48(1):10-16. [3] Uecker et al., ISMRM 2015 p2486. [4] Feng et al., MRM 2016:75:775-788.