Real-Time Golden Angle Radial iSSFP for Interventional MRI
Samantha Mikaiel1,2, Thomas Boyd Martin1,2, Kyung Sung1,2, and Holden H Wu1,2

1Radiological Sciences, University of California, Los Angeles, Los Angeles, CA, United States, 2Biomedical Physics, University of California, Los Angeles, Los Angeles, CA, United States

Synopsis

Real-time visualization is crucial to the success of MRI-guided minimally invasive cancer interventions. In this work we combine iSSFP with a golden-angle(GA) ordered radial trajectory and non-Cartesian parallel imaging to create a new real-time MRI sequence with good tissue contrast while suppressing bSSFP banding artifacts. Phantom and volunteer data were acquired and reconstructed using a combined sliding-window SPIRiT algorithm, at different frame rates, showing the capability of the sequence to achieve real-time imaging. These advantages of GA Radial iSSFP show its potential for improving real-time MRI-guided interventions.

Introduction

Real-time visualization is crucial to the success of MRI-guided minimally invasive cancer interventions, especially in the abdomen. The real-time MRI sequence must be ultrafast, while maintaining high soft-tissue contrast. RF-spoiled gradient echo (spGRE) is able to achieve real-time imaging speeds, but the SNR and tissue contrast is diminished. Balanced-SSFP (bSSFP) has better T2/T1 contrast and SNR, but suffers from banding artifacts(1). Half-Fourier acquisition single-shot turbo spin-echo (HASTE) can provide excellent T1 or T2 contrast, but only reaches low frame rates and is challenged by SAR restrictions. Compared to these sequences, integrated-SSFP (iSSFP)(2) is a special case of steady-state MRI that can suppress banding artifacts associated with bSSFP while maintaining similar T2/T1 contrast. In this work, we combine iSSFP with a golden-angle-ordered radial trajectory and non-Cartesian parallel imaging to create a new real-time MRI sequence with good tissue contrast while reducing banding artifacts.

Methods

[Real-time Golden-Angle Radial iSSFP Sequence] To create a radial iSSFP sequence, the slice select gradient of the bSSFP sequence was modified to achieve 2pi/voxel gradient spoiling (Fig1). The radial readouts were continuously acquired with golden-angle (GA) ordering(3).

[Reconstruction] The GA radial iSSFP data was reconstructed using a combined sliding-window SPIRiT(4) algorithm (Fig2). The sliding window reconstruction selects N spokes (temporal footprint N*TR) for each image frame and then continuously “slides” forward by M spokes (temporal resolution M*TR) until all of the data is reconstructed. For SPIRiT reconstruction, a calibration kernel is taken from a fully sampled central region of k-space for each image frame.

[Experiment Setup] All experiments were performed on a 3.0T MRI scanner (Prisma, Siemens) with 400-mm field-of-view, 192x192 matrix (unless noted), and 5-mm slices. A contrast phantom with different T1 and T2 compartments was placed in a waterbath, and imaged using spGRE (TE/TR=1.48/3 ms, FA=20o), bSSFP and iSSFP (TE/TR=1.9/3.9 ms, FA=70o), and HASTE (TE=68 ms, 256x256 matrix, 50% partial Fourier). This was first done with a Cartesian trajectory to investigate image contrast. Subsequently, GA radial versions of bSSFP and iSSFP were acquired. To demonstrate in vivo real-time imaging performance, a healthy volunteer was scanned during free breathing (FB) using the GA radial iSSFP and bSSFP sequences (TE/TR=1.9/3.9 ms, FA=41o). For comparison, FB GA radial GRE and breath-held (BH) Cartesian HASTE were also acquired in vivo.

[Imaging Analysis] The phantom images were visually examined to evaluate image quality, contrast and banding artifacts. The in vivo images were inspected for banding artifacts and tissue contrast differences.

Results

[Phantom Experiments] The Cartesian phantom images (Fig3) demonstrate the lack of T2 contrast in the spGRE image, as well as the banding artifacts in the bSSFP image. iSSFP has T2/T1 contrast similar to bSSFP, without any banding artifacts. For the GA radial bSSFP and iSSFP scans, it can be seen that there is similarly high T2/T1 contrast in both images, with iSSFP again suppressing the bSSFP banding artifacts.

[Volunteer Experiments] For the dynamic free-breathing scans, the N and M values for the sliding-window SPIRiT reconstructions were 144 and 20 respectively, which achieves a 576-ms temporal footprint (true frame rate 1.7 frames per second, fps) and a 80-ms temporal resolution (12-fps display frame rate). The data was also reconstructed with N = 55 and M = 20, achieving a 215-ms temporal footprint (true frame rate 4.5 fps) and 12-fps display frame rate. It was observed that iSSFP has contrast similar to the bSSFP acquisition without banding artifacts (Fig 4). By leveraging the GA radial acquisition pattern, volunteer motion was reconstructed and resolved with different frame rates (Fig5).

Discussion and Conclusion

Our GA radial iSSFP sequence combines the benefits of the iSSFP sequence with flexible dynamic GA radial acquisition. In phantom and volunteer scans at 3.0T, it is seen that GA radial iSSFP achieves good tissue contrast similar to bSSFP while suppressing banding artifacts. Sliding-window SPIRiT reconstruction allows dynamic visualization of subject motion with true frame rates from 1.7 to up to 4.5fps. Higher frame rates may be possible with additional constrained reconstruction. These advantages of GA Radial iSSFP show its potential for improving real-time MRI-guided interventions.

Acknowledgements

NSF Graduate Research Fellowship Program

References

1.Hargreaves BA. J. Magn. Reson. Imaging 2012;36:1300–13.

2. Martin TB, et al., Proc. Intl. Soc. Mag. Reson. Med. 2015;23:4157.

3. Winkelmann S, et al., IEEE Trans. Med. Imaging 2007;26:68–76.

4. Lustig M, et al., Magn. Reson. Med. 2010;64:457–71.

Figures

Figure 1: (a) depicts a standard bSSFP sequence. The slice select gradient was then adjusted to result in dephasing of 2pi/voxel to form the iSSFP sequence (b). The blue boxes highlight the new gradients.

Figure 2: Data is continuously acquired using GA radial iSSFP. Once a desired number (N) of spokes are selected, a calibration kernel is calculated from the center of k-space, and is used to reconstruct the image using SPIRiT. A shift of M spokes is performed to achieve sliding window selection.

Figure 3: Phantom images from the Cartesian (left) and GA Radial (right) scans. The Cartesian images are all set to the same widow level. The GA Radial data was reconstructed offline without coil intensity correction. All iSSFP images show similar contrast to bSSFP and do not have banding artifacts.

Figure 4: Single frame from the real time scans in both the Coronal and Sagittal planes. The image contrast in the stomach (red arrow), blood (white arrow), and the fat (arrowhead) are similar for bSSFP and iSSFP and different for spGRE. Banding artifacts are visible in only the bSSFP images.

Figure 5: Multiple frames of the FB GA Radial iSSFP Coronal scan at different reconstructed temporal footprints (frame rate) along with x-t plots at the x position marked by the red line.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
3579