Feasibility of brain intra-axonal microstructure imaging with ultrahigh B-encoding using MAGNUS ultra-high-performance gradients

The MAGNUS high-performance MRI gradient platform delivers G_max = 200–300 mT/m, and SR_max = 500–750 T/m/s using standard clinical 3.0T system power electronics. This enables the exploration of an expanded diffusion parameter space (b~7–≥30 ms/μm²) with reasonable SNR, along with substantially shorter diffusion encoding pulse-widths, echo times, reduced distortion, and blurring from shorter echo spacing. The choice of high b-value diffusion-encoding space can effectively suppress contributions from extra-axonal water, allowing for simplified biophysical models to be explored for non-invasive mapping of intra-axonal content. In this study, the feasibility and reproducibility of mapping in-vivo whole-brain effective intra-axonal radius (r_eff), using MAGNUS was assessed. By making use of a test-retest paradigm, reproducibility and sensitivity were evaluated for this new biomarker. Six healthy volunteers were imaged, after obtaining written informed consent, under local IRB-approved protocols with a focus on utilizing the maximum gradient strength of 300 mT/m. Multi-shell dMRI protocols, with a lower bound b = 7 ms/μm² were used for feasibility analysis and short (same-day) and long-term (7-days) test-retest repeatability. To aid in increased precision, a framework for rigorous post-processing incorporating real-valued diffusion data handling and gradient non-linearity correction was integrated. At 300 mT/m, simulations highlight a lower bound threshold for robust detectability of r_eff >1.41 μm. The simulated distribution function was consistent with in-vivo measurements, where a mean r_eff = 2.75 ± 0.15 μm was observed for whole-brain white matter (WM) across all volunteers. Left-Right brain white matter asymmetry as a function of r_eff was noted with segmentations of well-reported parcels, such as the corpus callosum and corticospinal tract, demonstrating good agreement with prior literature. Data highlighted good repeatability in voxel-wise and parcel-based estimates for short- and long-term test-retest analysis. A mean coefficient of variance of 3.2% for WM parcels across all volunteers was noted, with a reproducibility coefficient of 0.16 μm (6.6%) highlighting a lack of systemic bias. This study reports on the feasibility of investigating r_eff using MAGNUS. The analysis of repeatability established the floor of changes in the brain that can be observed in studies leveraging r_eff as a neuroimaging biomarker for white matter integrity or for investigating neuroplastic processes in the brain.