Higher magnetic field-strength (B₀) leads to improved signal-to-noise ratio (SNR). However, a higher B₀-field also reduces the transverse relaxation time (T₂) in the brain, which incurs a SNR penalty on techniques that require long echo times. Imaging of microscopic diffusion anisotropy is an example of such a technique, since it depends on the use of strong gradients with long durations to yield sufficient diffusion encoding [1–4]. Assuming values of T₂ in white matter of 70 and 46 ms at 3T and 7T [5], respectively, the ‘break-even’ echo time is approximately 110 ms. At longer echo times, the benefit of ultra-high fields is debatable.

The aim of this work was to investigate the feasibility of microscopic diffusion anisotropy imaging at 7T. In order to minimize the required echo time, we implemented asymmetric gradient waveforms, optimized in the framework suggested by Sjögren et al. [6].

Diffusion-weighted imaging was performed in two healthy volunteers (male, 31 years old) on a Philips Achieva 7T (60 mT/m, 100 T/m/s, 32-channel head coil) and on a Siemens Prisma 3T (80 mT/m, 200 T/m/s, 20-channel head coil). Diffusion encoding was performed in a spin-echo sequence with SPIR fat suppression and EPI-readout. Imaging parameters at 7T were TR=4000 ms, TE=103 ms, SENSE=3, partial Fourier=0.6, resolution=2×2×4 mm³, slices=9, b=100, 500, 1000, 1500 and 2000 s/mm², resulting in a total scan time of 14 min. All relevant parameters were approximately equal on the 3T apart from three parameters: TR=3000 ms, TE=100 ms, GRAPPA=2. Directional encoding was performed in 20 directions, and the isotropic encoding was repeated 20 times, for each b-value. Diffusion encoding was achieved by modulating the gradient amplitude according to Fig. 1, where the waveform is asymmetrical around the refocusing pulse (i.e., the shape is not identical before/after refocusing), and optimized to yield effective directional and isotopic encoding during 43 and 37 ms (before/after refocusing) [6]. This is unlike previous implementations based on magic angle spinning of the q-vector (qMAS) [7], where a symmetrical encoding design was used [1,3]. All images were corrected for motion and eddy-currents prior to analysis [8]. Parameter maps were calculated according to the framework suggested by Lasic et al. [1]. Briefly, KT is the total kurtosis observed in the signal, according to KT=3·var(D)/MD2, where var(D) is the variance of the underlying distribution of apparent diffusion coefficients, and MD is the mean diffusivity [1,9]. The isotropic and anisotropic kurtosis (KI and KA) are components of KT that are due to ‘heterogeneous isotropic components’ and ‘microscopic anisotropy’, respectively. Finally, the microscopic fractional anisotropy was calculated according to µFA=(2/3+4/45/KA)^–1/2, which may be interpreted as the FA that would be observed in the tissue if all tissue components were parallel [1,3]. The voxel-wise SNR was estimated from the isotopically encoded images for b=2000 s/mm² (n = 20).Figure 2 shows that parameter maps from 3T and 7T systems are qualitatively comparable, although the present 7T protocol appears more sensitive to imaging artifacts. The kurtosis maps in particular showed a substantial dependency on field strength, probably due to difference in relaxation times in CSF. Figures 3 and 4 show the signal-vs-b curves measured on the 7T system in ROIs placed in the corticospinal tract and in the lateral ventricles. Figure 5 shows the calculated voxel-wise SNR for b=2000 s/mm² at 7T in a single axial slice of the brain where the average SNR was approximately 8, which is sufficient for accurate quantification of the microscopic anisotropy.

To our knowledge, we present the first example of microscopic diffusion anisotropy imaging at a 7T human MRI system. We demonstrate that it is feasible to reach echo times that are short enough to benefit from the higher signal at the 7T platform, although additional optimization is warranted to further motivate the move to 7T. In this project, the use of asymmetrical waveforms were instrumental since a corresponding isotropic encoding by qMAS would increase the required echo time from 103 of 155 ms, translating into a signal loss of approximately 70% in white matter. Moreover, we note that image quality at 7T was adversely affected by fat artifacts and geometrical distortions. However, the current implementation is a work in progress, and we expect that such effects can be alleviated.

In conclusion, we demonstrated that imaging of microscopic anisotropy is feasible at 7T provided that asymmetrical waveforms are employed for the diffusion encoding.