Diffusion weighted imaging is a powerful tool for investigating changes resulting from spinal cord injury (SCI). Many models have been derived to characterize the various tissue components present in the signal in order to more accurately identify injury processes. This work compares measurements of several of these models in an in-vivo model of rat spinal cord injury at multiple time points to demonstrate the ability of each to detect injury.

A weight drop spinal cord contusion injury was induced at the T10 vertebral level in rats for severe (n=4), moderate (n=3), mild (n=3), and sham (n=4) injuries. Two in-vivo scanning sessions occurred post-injury: 2 days (acute) and 30 days (chronic).

Standard PGSE diffusion weighted images were acquired using a Bruker 9.4T Biospec Imaging system with a commercial 4-channel surface coil array (Bruker). A respiratory-gated EPI sequence was used with a TR/TE of >1500 / 28 ms. 30 diffusion directions were acquired with a gradient duration/separation of 8.25 / 12.5 ms for 3 b-values (500, 1000, 2000 s/mm²) and 15 B₀ averages. 12 axial slices were collected with an in-plane resolution of 0.20 mm² (128x128) and a slice thickness of 1.0 mm, which took approximately 65 minutes to collect. A sagittal FLASH sequence was used to center the slices over the T10 injury epicenter.

Images were corrected for motion and eddy currents using the Spinal Cord Toolbox¹. Custom Matlab routines were used for diffusion modeling. Standard Diffusion Tensor Imaging² (DTI) analysis was used to derive metrics of mean diffusivity (MD_DTI), axial diffusivity (AD_DTI), radial diffusivity (RD_DTI), and fractional anisotropy (FA_DTI). These metrics were further utilized with the Free Water Estimation³ (FWE) model to calculate corrected values of MD_FWE, AD_FWE, RD_FWE, and FA_FWE through removal of the free water fraction (FWF). Diffusion kurtosis fitting⁴ was implemented to calculate diffusivity parameters with corresponding values of mean, axial, and radial kurtosis. These were applied in the White Matter Tract Integrity⁵ (WMTI) model to determine the axonal water fraction (AWF), axonal diffusivity (D_a), extracellular parallel diffusivity (D_Epar), and extracellular perpendicular diffusivity (D_Eperp). Parameter means were taken from the 6 slices centered over the lesion epicenter using an automated, region-growing method in Matlab to derive whole-cord masks for each slice. Automated masks were examined for accuracy, with failed slices being excluded from the analysis (total of 8 slices out of 84).

Representative parameter maps from each model are given in Figure 1. Means obtained from automated ROIs closely matched those from hand-drawn masks (not shown).

Figure 2 shows a comparison of axial and radial diffusivity metrics at the acute and chronic time points. While no trends with injury severity were evident in the acute period, chronic diffusivity values showed that AD_FWE scales roughly with injury severity. This same trend was seen in D_a from the WMTI model. On the other hand, RD_DTI showed better separation of injury groups before FWE correction. D_Eperp from WMTI showed the strongest separation of injury groups.

FA measurements with FWE correction (Fig. 3) demonstrated better association with injury severity in both acute and chronic data. While an overall trend of increasing FWF is seen in the chronic data, AWF shows more consistency in its change with injury severity through both time points.

Edema and inflammation are major factors in SCI that alter volume fractions within tissues and lead to diffusivity changes. Parameters such as RD_DTI, and D_Eperp demonstrate increased diffusivity values in injury likely associated with the increased extracellular volume⁶. The decreased sensitivity of the RD_FWE metric reflects the removal of the edema component that serves as an injury marker. Free water removal does improve separation of injury groups for AD and FA, but the more advanced WMTI model shows stronger relationships with injury severity in the metrics of D_a, D_Eperp, and AWF. Thus while FWE can be used to improve DTI modeling, the kurtosis-based metrics demonstrate better potential for separating injury severity. This is especially true in the chronic time point, where inflammatory and glial responses to injury are more extensively developed following injury.Detection of injury severity with diffusion metrics shows that more advanced kurtosis-based parameters can better separate injury groups in a rat model of SCI. Metrics related to increased extracellular water show the strongest relationship with injury severity, especially in the chronic period, demonstrating the importance of edema and inflammation in driving diffusion-sensitive changes following injury.