Synopsis

The Gradient System Transfer Function (GSTF) was used to implement a fully automatic gradient pre-emphasis, enabling double-oblique MR imaging for arbitrary k-space trajectories. The developed method was tested using a standard 2D FLASH spiral sequence with several orientations in a structural phantom and an exemplary in vivo cardiac study.

Purpose

When applying varying magnetic field gradients in an MR sequence, the determination and application of a Gradient System Transfer Function (GSTF) can compensate imperfections of the magnetic field dynamics. This reduces k-space deviations and ultimately image artifacts. Recently a prototype GSTF pre-emphasis was proposed [1].

The purpose of the work presented here was to extend the prototype towards a fully automated pre-emphasis on all three gradient axes that enables optimized imaging for arbitrary k-space trajectories and double-oblique slice orientations.

Methods

A phantom based technique [1,2,3] was used to determine the GSTF self-terms of all three axes of the gradient system of a 3T MR scanner (MAGNETOM Prisma, Siemens Healthcare GmbH, Erlangen, Germany). 12 triangular input gradients were played out and phase evolutions were measured in two parallel slices equidistantly from the scanner isocenter, for the physical x-, y- and z-axis, respectively (TR = 1 s, slice thickness = 3 mm, slice distance from isocenter = ±16.5 mm, flip-angle = 90°, dwell-time = 8.7 µs, 100 measurements). For arbitrary slice orientations, the concept of the automated pre-emphasis was to project any gradient waveform prepared by the pulse sequence onto the physical x-, y- and z-axis of the scanner, followed by a convolution with the inverse GSTF determined for the respective dimension.

The technique was first tested by imaging a structural phantom in several orientations using a 2D FLASH spiral sequence. The following parameters were applied: spiral angle increment ϕ ≈ 222.49°, TR = 11.5 ms, TE = 3.3 ms, flip-angle = 10°, readout BW = 250 Hz/px, in-plane spatial resolution ≈ 1 mm, slice thickness = 3.5 mm and 256 spiral interleaves. All datasets were reconstructed using convolution gridding as described in [4].

Subsequently, an in vivo cardiac spiral cine study was performed using the same sequence, except for an increased slice thickness (8 mm) and in-plane spatial resolution (≈ 2 mm). 791 consecutive spiral readouts were acquired within a breath-hold of a healthy volunteer. Using self-gating, data were then sorted into 20 heart-phases (temporal resolution < 45 ms) and reconstructed by the compressed sensing technique presented in [5]. For comparison, both the phantom and the in vivo acquisition were repeated with pre-emphasis switched off.

Results

In Fig. 1, three acquisitions of the structural phantom in transversal and double-oblique orientations are shown with and without applied pre-emphasis. All data sets were reconstructed with the nominal trajectory, leading to high image quality when the pre-emphasis was applied. In contrast, typical artifacts caused by trajectory errors are visible in the images without pre-emphasis.

Fig. 2 shows the results of the cardiac in vivo study. Again, image artifacts are present for the acquisition without pre-emphasis (see arrows). The supplementary video of the entire cine series with applied pre-emphasis (Fig. 3) exhibits an image quality similar to clinical acquisitions based on fully sampled Cartesian imaging. At most, a slight jitter remains in the liver region.

Discussion & Conclusion

Phantom and in vivo measurements demonstrate an apparent enhancement of image quality when considering the implemented automatic pre-emphasis. Especially cardiac applications benefit from the proposed method, as the position of the heart typically requires double-oblique scanning.

In summary, the presented study illustrates the substantial advantage of considering the GSTF-based compensation of gradient inaccuracies already within the pulse sequence code, i.e. before playing out any gradient waveform. The developed strategy facilitates double-oblique scanning for arbitrary k-space trajectories by minimizing artifacts typically present in investigations based on non-Cartesian imaging.

Funding

Acknowledgements

References

[1] Stich M. et al., Gradient waveform pre-emphasis based on the gradient system transfer function, Magnetic Resonance in Medicine, 80(4):1521-1532 (2018)

[2] Liu H. et al., Accurate Measurement of Magnetic Resonance Imaging Gradient Characteristics, Materials, 7:1-15 (2014)

[3] Campbell-Washburn A. E. et al., Real-Time Distortion Correction of Spiral and Echo Planar Images Using the Gradient System Impulse Response Function, Magnetic Resonance in Medicine, 75(6):2278-2285 (2016)

[4] Fessler J. A. et al., Nonuniform Fast Fourier Transforms Using Min-Max Interpolation, IEEE Transactions On Signal Processing, 51(2):560-574 (2003)

[5] Otazo M. et al., Low-rank plus sparse matrix decomposition for accelerated dynamic MRI with separation of background and dynamic components, Magnetic Resonance in Medicine, 73(3):1125-1136 (2015)