Background

NEXI¹ (or SMEX²) is a two-compartment diffusion model of gray matter (GM) microstructure that accounts for inter-compartment exchange. Its parameters are the inter-compartment exchange time t_ex, the intra and extra-neurite apparent diffusivities D_i and D_e and the intra-neurite signal fraction f. The narrow-pulse approximation (δ→0) enables an analytical fit using, e.g. standard non-linear least squares (NLS), and is a valid approximation for data acquired on preclinical scanners, on which NEXI was first demonstrated in vivo¹ and ex vivo². Accounting for the actual wide diffusion pulses in the model equations has significant relevance for data acquired on clinical scanners. Here, we aimed to test the performance of NEXI in the human cortex in vivo on a clinical MRI scanner. We compared parameter estimates obtained using the Narrow Pulse Approximation (NEXI_NPA)¹ or the Actual Wide Pulses (NEXI_AWP)² to literature values, and related t_ex estimates with the myelin water fraction obtained by multicomponent T₂ relaxation analysis³, assuming the exchange time mirrors permeability, and is related to myelination.

Methods

Clinical data: Six healthy volunteers (3M, 26.5±1.1 years old) were scanned on a 3T system with 80 mT/m gradients (MAGNETOM Prisma, Siemens Healthcare, Erlangen, Germany), five of whom were rescanned on a different day. Acquisition: An MPRAGE was acquired for anatomical reference (1-mm isotropic resolution). Diffusion-weighted images were acquired using a PGSE-EPI research sequence with the following parameters: b-values=1.00, 2.00, 3.20, 4.44 and 5.00ms/µm², 20 directions per shell, diffusion times Δ=28.3, 36.0, 45.0, 55.0 and 65.0ms, δ=16.5ms, 4 b=0 images, spatial resolution=2x2x2 mm³, total scan time: 35min. Multi-echo T₂ data were collected using a 3D multi-echo accelerated gradient and spin echo (GRASE) research application sequence⁴ with voxel-size=1.8x1.8x1.8mm³; ΔTE/N-echoes=10.94ms/32; TR=1s, N-slices=76.
Processing: Multi-shell multi-diffusion time data were preprocessed jointly; steps included MP-PCA magnitude denoising⁵, Gibbs ringing⁶, distortion and Eddy current corrections⁷. The gray matter region of interests (ROIs) from the Desikan-Killiany-Tourville atlas⁸ were segmented on the MPRAGE image using FastSurfer⁹ and projected onto the diffusion native space using linear registration¹⁰. Parametric maps of NEXI_NPA or NEXI_AWP, both corrected for Rician Mean¹¹, were estimated from the powder-average signals using non-linear least-squares. Myelin Water Fractions (MWF) were estimated using the L-curve-I non-parametric T₂ relaxometry method for myelin water quantification³. The spatial distribution of NEXI features was also examined using inflated brain surfaces obtained using Connectome Workbench¹².

Results and discussion

Despite δ being over half of the shortest Δ, NEXI_NPA and NEXI_AWP yielded reasonably similar estimates (Fig. 1): t_ex ranged 10-30ms, in line with estimates in human gray matter^11,13,14 and rat cortex in vivo¹, particularly for NEXI_AWP. Neurite fraction ranged 0.42-0.46, was the most robust across models and similar to that obtained in humans on a Connectom scanner¹¹. D_i was lower with NEXI_AWP and overall more biologically plausible^15,16, as the NEXI_NPA estimate exceeds the value in free water. The two versions of NEXI show almost identical patterns across the cortical surface (Fig. 2), in particular higher f in the visual cortex, but also and especially in the motor cortex, which was less apparent in a previous study on Connectom¹¹. Importantly, these NEXI implementations on clinical data retain sensitivity to individual differences and high intra-subject reproducibility (Fig. 3). However, NEXI_AWP parameter estimates have a lower variance than NEXI_NPA ones (Fig. 1). This smaller variance also reduces NEXI_NPA sensitivity to individual differences, mainly for f (Fig. 3). Finally, we report a significant positive correlation (NEXI_NPA: r=.73, p<.0001; NEXI_AWP: r=.67, p<.0001) between the mean t_ex and the mean MWF estimated in each cortical ROI (Fig. 4). This agreement suggests the exchange time estimated using NEXI is a reasonable proxy for membrane permeability, which is expected to decrease with myelination. A significant positive correlation with MWF was also found for D_e (NEXI_NPA: r=.49, p<.0001; NEXI_AWP: r=.68, p<.0001), which may be consistent with myelinated axons/dendrites showing a more coherent organization and faster overall extra-cellular diffusivity. Correlations between MWF and f or D_i were non-significant. It remains, however, to be determined how populations with different NEXI parameter values within a voxel would be reflected by a single exchange time estimate.

Conclusion

NEXI was successfully estimated in the human cortex on data acquired on a clinical scanner. Results are consistent with previous findings in human gray matter and from in vivo imaging in the rat cortex and display previously observed patterns. The NEXI_AWP version, which includes wide diffusion pulses, yields more robust and biologically plausible results, but perhaps with challenging sensitivity to individual differences. Strong correlation of t_ex with the Myelin Water Fraction is consistent with expected relationship between cell membrane permeability and myelination.

Acknowledgements

QU, TP and IJ are supported by SNSF Eccellenza grant PCEFP2_194260.

References

[1] Jelescu, NeuroImage 2022 [2] Olesen, NeuroImage 2022 [3] Canales-Rodríguez, Medical Image Analysis 2021 [4] Piredda, MRM 2020 [5] Veerart, NeuroImage 2016 [6] Kellner, MRM 2016 [7] Andersson, NeuroImage 2016 [8] Klein and Tourville, Frontiers in Neuroscience 2012 [9] Henschel, NeuroImage 2020 [10] Avants et al., Insight j, 2009 [11] Uhl, Arxiv 2023 [12] Marcus, Frontiers in Neuroinformatics 2011 [13] Lee, NeuroImage 2020, [14] Veraart, eLife 2020 [15] Dhital, NeuroImage 2019 [16] Howard, NeuroImage 2022