Diffusion tensor imaging (DTI) is a common technique for evaluating microstructural changes in spinal cord injury (SCI), but cannot adequately resolve diffusion markers of axonal integrity, the most direct marker of functional status¹, from confounding changes related to edema and inflammation². While more advanced models can isolate the axonal component^3,4,5, imaging duration and post-processing requirements limit clinical feasibility. A double diffusion encoding (DDE) scheme has been devised specifically to evaluate axonal integrity following SCI while reducing scan duration and minimizing post processing. In-vivo comparison of standard DTI metrics with DDE parameters was performed in a rat model of SCI.

All diffusion data were collected using a Bruker 9.4T Biospec imaging system. DTI images were acquired with a standard PGSE sequence with gradient duration (δ) of 8.25 ms and separation (Δ) of 12.5 ms, using a 4-shot, respiratory gated EPI sequence (TR/TE = 1500/28 ms) consisting of 30 directions and 3 b-values (500, 1000, and 2000 s/mm²). Images had an in-plane resolution of 0.20 mm² (128x128) and a slice thickness of 1.0 mm. Acquisition time was approximately 65 minutes.

DDE was implemented with two pairs of Stejskal-Tanner gradients (Fig. 1) with δ/Δ = 12/6 ms for each gradient pair. The first gradient pair (“filter”) was applied perpendicular to the spinal cord to attenuate non-neuronal tissue signal. The second diffusion weighted direction (“probe”) was aligned parallel to the spinal cord to sample axial diffusion. An EPI sequence implementation (TR/TE = 1750/51.07 ms) was performed using a b=4000 s/mm² filter and 5 probe b-values from 0-1000 s/mm² (20 minute acquisition). Another implementation using a Point RESolved Spectroscopy (PRESS) sequence (TR/TE = 3000/42.26 ms) was developed to acquire a single 10x10x6 mm³ voxel with a b=2000 s/mm² filter and 9 probe b-values ranging from 0-2000 s/mm². Acquisition time was approximately 3 minutes.

In naïve rats, DTI and DDE-EPI sequences were acquired with a quadrature Litz coil (Doty Scientific) in the cervical spinal cord (n=3). For the rat injury model, a weight drop contusion at the T10 vertebral level was performed with severe, moderate, mild, and sham (control) severities (n=14). At 48 hours post-injury, DTI and DDE-PRESS diffusion data were acquired at the lesion epicenter with a 4-channel surface coil array (Bruker).

Six DTI slices covering the extent of the DDE voxel were used to compare injury diffusion metrics. Images were motion and eddy current corrected, then DTI parameters calculated using weighted linear least squares fitting. Whole-cord regions of interest were used to obtain Fractional Anisotropy (FA), Mean Diffusivity (MD), Axial Diffusivity (AD), and Radial Diffusivity (RD) averages. For DDE data, the integral of the magnitude data was fit using a monoexponetial model to derive axial diffusivity (AD_DDE) while a biexponential fit that included a slower restricted diffusion term was used to derive slow diffusivity (D_slow) and the signal fraction with this restricted diffusion (f_R). Injury effect analyzed with ANOVA and regression analyses.

Analysis of signal decay in tissues surrounding the spinal cord demonstrated that non-neural tissue is attenuated with a b-value >2000 s/mm² (not shown). The filter did not significantly alter AD values in a naïve rat (Fig. 2) and showed no difference between the EPI and PRESS acquisition methods (Fig. 3).

ANOVA to detect the main injury effect showed no significance in the DTI parameters at 48 hours, while DDE-derived AD_DDE and f_R showed a significant effect (Fig.4). Regression analysis of injury biomechanics (Fig. 5) and functional testing (not shown) showed similar significance in DDE metrics while DTI measurements showed no significance.

The alignment of fibers in the spinal cord allows attenuation of non-neuronal signal while preserving that of the spinal cord using a strong perpendicular diffusion weighting. By combining this with multiple parallel diffusion weightings, estimates of AD can be achieved using a reduced number of directions that in a healthy, naïve rat agree with those achieved using standard DTI methods. Furthermore, as signal of non-interest is removed, a spectroscopic readout allows a whole-cord calculation of diffusivity values without involved and subjective ROI analysis.

The closer association of DDE metrics with injury severity is thought to be related to the attenuation of extracellular fluid (edema), thus resulting in signal with a greater fraction of the axonal component. Functional analyses also show significant association with these measurements, demonstrating potential as a rapid injury assessment tool.

DDE applied to SCI provides a substantial decrease in acquisition time while improving the detection of injury severity in a rat model. This has potential applications for a fast determination of injury severity in preclinical or clinical applications.