Diffusion Across Field Strengths
Mara Cercignani1
1Cardiff University, United Kingdom

Synopsis

Keywords: Contrast mechanisms: Diffusion

This lecture will discuss the principles and challenges of diffusion MRI (dMRI) at different field strengths. Low field (<1T) poses signal-to-noise ratio challenges, while high field (>3T) faces issues like field inhomogeneity and shorter T2 relaxation times. At high field, strategies to minimize SNR loss include non-Cartesian k-space trajectories and 3D sequences. Conversely, at low field, long scan times hinder data quality, necessitating innovative acquisition and image reconstruction techniques. The talk will discuss the attractiveness and challenges of different field strengths and proposes strategies for addressing dMRI difficulties.

Introduction

Diffusion MRI (dMRI) probes tissue microstructure by exploiting the random motion of water molecules within tissue. Magnetic field gradients are typically used to encode molecular displacement into signal attenuation. This means that as the sensitivity to diffusion increases, the signal to noise ratio (SNR) decreases. In addition, dMRI typically require long echo times (TE) which also contribute to reduce the SNR. As SNR is assumed to vary linearly with field strength, these observations would suggest that low field (<1T) might pose a significant challenge to dMRI. On the other hand, the nominal advantage of high field (>3T) is partially outweighed by other factors (B0 and B1 field inhomogeneity, shorter T2, etc). This means that at both, high and low field, dMRI requires some ‘thinking outside the box’, rather than the direct replication of protocols that work well in the B0 range 1T-3T.This talk will discuss the challenges of doing dMRI at low and high field, and will provide an overview of some of the proposed solutions.

Basics of Diffusion MRI

The most common acquisition sequence for dMRI is the so-called pulsed-gradient spin-echo (PGSE 1) echo planar imaging (EPI 2), i.e. a spin-echo prepared EPI, with the addition of 2 large gradients either side of the refocusing pulse. The amount of diffusion-weighting (b factor or b value 3) is determined by the amplitude and the duration/separation of the gradients. As the former property is hardware-limited, PGSE sequences tend to use long echo times (TEs) to accommodate the required diffusion gradients. The EPI readout, which minimises the effects of involuntary motion, makes this acquisition sensitive to magnetic susceptibility, which causes geometric distortions4, as well as other off-resonance artifacts such as chemical-shift. Due to diffusion anisotropy in the white matter5, usually more than one diffusion-weighted volume is acquired, making scan time an important variable in the implementation of dMRI.

Diffusion MRI at high field

Based on the rough approximation of a linear relationship between SNR and field strength6, an increase in SNR of approximately 2.3 is expected at 7T compared to 3T. However this is typically not achieved because T2 tends to decrease with B0. This means that long TE acquisitions might actually provide less SNR at 7T than at 3T. This is further exacerbated by the increased magnetic susceptibility, which calls for larger bandwidth at 7T to reduce distortions, thus further reducing the SNR. Strategies to minimise the TE include non Cartesian k-space trajectories such as spirals7 or the use of stimulated echo sequences8. One of the attractions of high field is the potential for higher resolution; however, this is difficult to accomplish with standard 2D EPI, and 3D sequences might be a better choice for these purposes9. Other considerations for diffusion MRI at high field include the increased power deposition which may limit applicability and the increased inhomogeneity of both B1 and B0 fields. These and other potential challenges are covered in excellent reviews (e.g., 10). Diffusion-weighted spectroscopy11, which measures the diffusion properties of intracellular metabolites such as NAA and Choline, might benefit from higher field strength more than dMRI, thanks to the increased spectral resolution.

Diffusion MRI at low field

By contrast, at low field strengths, dMRI is challenging primarily because of the lack of signal. As SNR and acquisition time are usually inversely correlated, this often results in long scan times at low field to ensure sufficient data quality. This requirement can be unacceptable for non-compliant populations such as for example infants or certain clinical groups. The long acquisition time also prevents the use of many diffusion encoding directions, effectively limiting the applicability of high order diffusion models12. Low field MRI ranges from 10 mT to 1 T and typically includes imaging systems with very different characteristics. At the lower end of this magnetic field range, portable and point-of-care systems are found. These scanners which typically use permanent magnets and no chillers or shielding pose additional challenges for diffusion MRI. They generally suffer from very poor B0 field homogeneity and provide low gradient amplitude. Despite these limitations, examples of diffusion tensor MRI have been provided. Generally, non-EPI sequences are preferred (due to the challenges with switching the gradients rapidly). Due to the minimal SAR constraints, 3D RARE readouts are particularly suitable, although they require proper navigation, when combined with diffusion encoding13. For quantitative qMRI, gradient non-uniformity is also a concern. In the range of field strength above 0.1 T, we typically find superconducting systems. In this case, losing the attractive option of portability is compensated by better field homogeneity and generally better image quality due to improved shielding. Higher gradient amplitude is also achievable, although the issue of Maxwell terms (concomitant fields) can be significant at low field strengths14. Concomitant gradient fields are generated in order to obey Maxwell equations every time a gradient is switched on. As their amplitude is proportional to the ratio between the gradient amplitude and the static field, they are considered negligible for most applications at high field. However, with diffusion encoding at low field, caution in the diffusion encoding design must be used to prevent their effect to completely destroy the signal. When working close to the noise level, clever image reconstruction and denoising techniques become very important, and the recent advances in machine learning image processing approaches find many areas of application in low field MRI in general, and dMRI in particular (e.g.,15,16).

Conclusions

Diffusion MRI has become one of the most popular MRI techniques in research as well as in certain clinical applications. UHF is an attractive option for its promise of higher SNR/improved resolution; however, UHF comes with new challenges that may outweigh its advantages for dMRI. By contrast, low field is attractive for its reduced cost and housing requirement, but the SNR penalty can significantly impact the quality of dMRI. Hardware, acquisition, and image analysis strategies that can offer interesting solutions for these problems will be discussed, while examples of successful implementation across the field strength will be shown.

Acknowledgements

No acknowledgement found.

References

  1. Tanner JE, Stejskal EO. Restricted Self-Diffusion of Protons in Colloidal Systems by Pulsed-Gradient Spin-Echo Method. J Chem Phys. 1968;49(4):1768-+. doi:Doi 10.1063/1.16703062.
  2. Mansfield P. Multi-Planar Image-Formation Using Nmr Spin Echoes. J Phys C Solid State. 1977;10(3):L55-L58. doi:Doi 10.1088/0022-3719/10/3/0043.
  3. Lebihan D, Breton E, Lallemand D, Grenier P, Cabanis E, Lavaljeantet M. Mr Imaging of Intravoxel Incoherent Motions - Application to Diffusion and Perfusion in Neurologic Disorders. Radiology. Nov 1986;161(2):401-407. doi:DOI 10.1148/radiology.161.2.37639094.
  4. Jezzard P, Balaban RS. Correction for geometric distortion in echo planar images from B0 field variations. Magn Reson Med. Jul 1995;34(1):65-73. doi:10.1002/mrm.19103401115.
  5. Moseley ME, Kucharczyk J, Asgari HS, Norman D. Anisotropy in diffusion-weighted MRI. Magn Reson Med. Jun 1991;19(2):321-6. doi:10.1002/mrm.19101902226.
  6. Hoult DI, Lauterbur PC. Sensitivity of the Zeugmatographic Experiment Involving Human Samples. J Magn Reson. 1979;34(2):425-433. doi:Doi 10.1016/0022-2364(79)90019-27.
  7. Engel M, Mueller L, Döring A, Afzali M, Jones DK. Maximizing SNR per unit time in diffusion MRI with multiband T-Hex spirals. Magnet Reson Med. Dec 29 2023;doi:10.1002/mrm.299538.
  8. Merboldt KD, Hanicke W, Frahm J. Diffusion imaging using stimulated echoes. Magn Reson Med. Jun 1991;19(2):233-9. doi:10.1002/mrm.19101902089.
  9. Wu W, Poser BA, Douaud G, et al. High-resolution diffusion MRI at 7T using a three-dimensional multi-slab acquisition. Neuroimage. Dec 2016;143:1-14. doi:10.1016/j.neuroimage.2016.08.05410.
  10. Gallichan D. Diffusion MRI of the human brain at ultra-high field (UHF): A review. Neuroimage. Mar 2018;168:172-180. doi:10.1016/j.neuroimage.2017.04.03711.
  11. Merboldt KD, Horstermann D, Hanicke W, Bruhn H, Frahm J. Molecular Self-Diffusion of Intracellular Metabolites in Rat-Brain Invivo Investigated by Localized Proton Nmr Diffusion Spectroscopy. Magnet Reson Med. Jan 1993;29(1):125-129. doi:DOI 10.1002/mrm.191029012312.
  12. Alexander DC, Dyrby TB, Nilsson M, Zhang H. Imaging brain microstructure with diffusion MRI: practicality and applications. Nmr Biomed. Apr 2019;32(4)doi:ARTN e384110.1002/nbm.384113.
  13. Trouard TP, Theilmann RJ, Altbach MI, Gmitro AF. High-resolution diffusion imaging with DIFRAD-FSE (diffusion-weighted radial acquisition of data with fast spin-echo) MRI. Magn Reson Med. Jul 1999;42(1):11-8. doi:10.1002/(sici)1522-2594(199907)42:1<11::aid-mrm3>3.0.co;2-j14.
  14. Bernstein MA, Zhou XJ, Polzin JA, et al. Concomitant gradient terms in phase contrast MR: analysis and correction. Magn Reson Med. Feb 1998;39(2):300-8. doi:10.1002/mrm.191039021815.
  15. Lin H, Figini M, D'Arco F, et al. Low-field magnetic resonance image enhancement via stochastic image quality transfer. Med Image Anal. Jul 2023;87:102807. doi:10.1016/j.media.2023.10280716.
  16. Zhang Y, He W, Chen F, Wu J, He Y, Xu Z. Denoise ultra-low-field 3D magnetic resonance images using a joint signal-image domain filter. J Magn Reson. Nov 2022;344:107319. doi:10.1016/j.jmr.2022.107319
Proc. Intl. Soc. Mag. Reson. Med. 32 (2024)