Oscillating gradient spin echo diffusion time effects implicate variations in neurite beading for the heterogeneous reduced diffusion in human acute ischemic stroke lesions

Abstract

Purpose: Monte Carlo simulations and short diffusion time measurements have suggested neurite beading and swelling as the underlying mechanism of reduced diffusion in acute stroke, although the observed diffusion time dependence is often heterogeneous and not yet fully understood. This study aimed to investigate the heterogeneity of diffusion time effects in ischemic lesions and explore the potential microstructural basis with Monte Carlo simulations.

Methods: Pulsed gradient spin echo (PGSE, diffusion time 40 ms) and oscillating gradient spin echo (OGSE 40 Hz, diffusion time ˜5.1 ms) were acquired within 5 min in 39 acute ischemic stroke patients at 3 T. Mean, axial, and radial diffusivity differences between OGSE and PGSE (ΔMD/ΔAD/ΔRD) were compared between lesion and contralateral tissues (white and gray matter). Monte Carlo diffusion simulations of beaded axons for the experimental waveforms were used to investigate the effects of neurite morphology on time-dependent diffusivity changes.

Results: PGSE yielded the typical mean diffusivity (MD) reduction of -40 ± 10% in ischemic lesions, whereas it was less at -29 ± 11% for OGSE 40 Hz. The OGSE-PGSE diffusion time difference was greater in lesions (ΔMD = 0.12 ± 0.06 × 10^-3 mm²/s) than contralateral white matter (ΔMD = 0.04 ± 0.06 × 10^-3 mm²/s), consistent with larger beading amplitude (0.18-0.43) and intracellular volume fraction (0.61-0.78) in lesions. ΔMD maps revealed regional variation with the largest effects in internal capsule and corona radiata.

Conclusion: This study found greater diffusion time effects in ischemic regions with purportedly larger axons. Monte Carlo simulations further support that the pronounced OGSE-PGSE diffusivity differences are expected in large axons with high beading amplitude.

Keywords: Monte Carlo diffusion simulation; neurite beading; oscillating gradient spin echo diffusion MRI; stroke.

FIGURE 1

Schematic representation of (A) diffusion‐sensitizing gradient waveforms and (B) the corresponding diffusion encoding power spectrum for pulsed gradient spin echo (PGSE) “0 Hz” and oscillating gradient spin echo (OGSE) “40 Hz” one‐period and “50 Hz” two‐period trapezoid cosines. The use of 1 period causes sizeable lower frequency harmonics below the desired 40 Hz frequencies. Note that the second half of gradients would be flipped when including a 180° RF refocusing pulse.

FIGURE 2

Raw pulsed gradient spin echo (PGSE) DWI b500, PGSE‐ and oscillating gradient spin echo (OGSE)‐derived mean diffusivity (MD), and the MD difference maps (along with the fractional anisotropy [FA] maps for anatomical information) for a single slice from seven representative acute stroke patients. Lesion white matter (WM) for all strokes is delineated by white outlines. The lower MD in the ischemic lesion is not as pronounced on OGSE as it is on PGSE, which is better visualized on the OGSE MD‐PGSE MD difference map. There is notable heterogeneity of OGSE‐PGSE MD difference in the lesions, with larger effects in the lesion WM.

FIGURE 3

Mean ± SD of mean diffusivity (MD), axial diffusivity (AD), radial diffusivity (RD), and fractional anisotropy (FA) values in contralateral (blue and gray) and lesion (orange and yellow) white matter (WM) (A–D) (n = 39) and gray matter (GM) (E–H) (n = 30). The percent reduction of MD, AD, and RD in lesion WM relative to contralateral WM was greater for typical pulsed gradient spin echo (PGSE) (37%–43%) than oscillating gradient spin echo (OGSE) (24%–34%). From PGSE to OGSE, MD, AD, and RD had larger increases in lesion WM (20%–27%) than in contralateral WM (4%–6%), with RD showing the greatest percent change. FA was lower in lesion WM relative to contralateral tissues. Similar trends of the DTI parameters were observed in GM, albeit to a less extent than in WM. (*p < 0.05).

FIGURE 4

Across the 39 acute stroke patients (each dot), the oscillating gradient spin echo (OGSE)‐pulsed gradient spin echo (PGSE) ΔMD had a larger range in the lesion white matter (WM) (0.02–0.26 × 10⁻³ mm²/s) than in contralateral WM (−0.01–0.08 × 10⁻³ mm²/s). Interestingly, patients with lesions located mainly in the internal capsule (IC) or corona radiata (CR) (red dots) yielded lesion OGSE‐PGSE ΔMD larger than 0.13 × 10⁻³ mm²/s, whereas lesions involving other WM tracts (green dots) mostly had lower ΔMD.

FIGURE 5

Mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD) (columns 1–3) changes are shown for pulsed gradient spin echo (PGSE) (solid line) and oscillating gradient spin echo (OGSE) (dash line) as a function of beading amplitude and intracellular volume fraction for cylinder (axon) radius of (A) 1, (B) 2, (C) 3 and (D) 4 μm. PGSE values for MD, AD, and RD are generally consistent across all radii examined, whereas AD decreases substantially with increased beading amplitude and RD decreases as intra‐axonal volume fraction decreases (0.5 red to 0.8 magenta). (A,B) In small radius axons, these diffusivities remain similar between PGSE and OGSE. (C,D) As axon radius increases, OGSE has greater deviations from PGSE. The reduction in AD with increased beading amplitude is less pronounced while there is a greater increase in RD with increased beading amplitude. On the contrary, the changes as a function of volume fraction become less pronounced with OGSE. These combined effects result in MD values for OGSE in larger radius axons being generally insensitive to beading amplitude and with only minor sensitivity to volume fraction changes. Free diffusivity = 2.1 × 10⁻³ mm²/s for demonstration. Note that the abrupt RD increase from beading amplitude 0.3 to 0.4 at a large volume fraction of 0.8 results from the local axon deformation to prevent overlap between adjacent beads.

FIGURE 6

Mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD) (columns 1–3) varying by beading amplitude (x axis) and volume fraction (y axis) from Monte Carlo diffusion simulations of (A) pulsed gradient spin echo (PGSE) and (B) oscillating gradient spin echo (OGSE). The subtraction of the simulated MD/AD/RD between OGSE and PGSE (ΔMD, ΔAD, ΔRD) are shown in (C). Free diffusivity and radius were selected as 2.1 × 10⁻³ mm²/s and 2.5 μm, respectively, for demonstration, based on the mean free diffusivity (2.08 ± 0.18 × 10⁻³ mm²/s) and radius (2.4 ± 0.3 μm) that best matched simulation and experimental ΔAD and ΔRD values from the 39 patients. The best‐fit of beading amplitude and volume fraction (mean ± SD across 39 patients) using ΔAD and ΔRD are demonstrated for lesion (white circle and error bar) and contralateral WM (black circle and error bar). In the healthy contralateral white matter (WM), this yielded an estimated beading amplitude of 0.18 ± 0.05 and volume fraction of 0.61 ± 0.09, whereas in the lesion WM, it yielded the higher beading amplitude of 0.43 ± 0.17 and larger volume fraction of 0.78 ± 0.09.

FIGURE 7

An illustration of the axon geometry shapes/packing that yield the best pattern matching results between Monte Carlo simulated and experimentally observed ΔAD and ΔRD in lesion and contralateral white matter (WM) for oscillating gradient spin echo (OGSE) and pulsed gradient spin echo (PGSE) in 39 stroke patients. It showed a simultaneous increase of both beading amplitude and intracellular volume fraction in lesion WM (right column) relative to contralateral WM (left column). The free diffusivity (2.1 × 10⁻³ mm²/s) and axon radius (2.5 μm) were kept constant.

FIGURE 8

Mean ± SD of (A) mean diffusivity (MD), (B) axial diffusivity (AD), and (C) radial diffusivity (RD) values from contralateral (blue) and lesion (orange) white matter (WM) (n = 7 stroke patients) for pulsed gradient spin echo (PGSE), oscillating gradient spin echo (OGSE) 40 Hz and OGSE 50 Hz using b value of 400 s/mm². The bolded numbers give % change in lesion WM relative to contralateral WM (no significant difference is denoted by n.s.). Lesion MD, AD, and RD are progressively less reduced (relative to contralateral WM) with shorter diffusion time. Lesion WM AD exhibits the largest reduction overall, whereas lesion RD reduction is not significant for the OGSE scans. The MD and AD difference between OGSE 40 Hz and 50 Hz is significant but small, whereas RD change in contralateral WM was not significant. (*p < 0.05).

FIGURE 9

DWI b400, pulsed gradient spin echo (PGSE)/oscillating gradient spin echo (OGSE) 40 Hz/OGSE 50 Hz mean diffusivity (MD) maps, and PGSE directionally encoded color fractional anisotropy (FA) maps are shown for ischemic lesions (arrow) in 3 acute ischemic stroke patients. (A) Patient 1: the lesion shows a large MD reduction on PGSE which is less, albeit still markedly lower, on both OGSE scans with only a small difference between 40 and 50 Hz. (B) Patient 2: the lesion has a smaller drop in PGSE MD, with a lower but persistent MD reduction on both OGSE MD maps. (C) Patient 3 shows an interesting observation where the lesion located in the right corticospinal tract even “disappears” on both OGSE MD maps.