Diffusional kurtosis is a measure of the non-Gaussianity of the diffusion process of water molecules in tissue.¹ DKI may be a more sensitive marker for pathophysiologic changes in cellular microstructure after acute ischemic stroke (AIS) than diffusion tensor imaging (DTI) metrics such as fractional anisotropy (FA).^{1, 2} Furthermore, in a subset of AIS patients who undergo successful revascularization treatment, lesions on acute trace diffusion-weighed MRI (DWI) have been noted to “reverse”.^{3, 4} Coupled with changes in diffusivity, DKI may provide complementary insight into the extent of neuronal injury.² We therefore investigated DKI metrics as a function of tissue outcome in AIS patients.DKI from 18 AIS patients enrolled in a prospective serial MRI study were analyzed. The first MRI was acquired for clinical purposes within 12 h from when the patient was last-known-well on a 1.5T GE scanner, and the second MRI prior to discharge on a 3T Siemens system. MR perfusion-weighted imaging (MRP) was acquired at both sessions with gradient echo echo-planar imaging (TR/TE=1500/40 ms at 1.5T and TR/TE=1500/35 ms at 3T, 80 timepoints) during the first pass of an intravenous bolus injection of a gadolinium-based contrast agent. Perfusion maps (CBF, CBV, MTT and Tmax) were calculated using deconvolution⁵ with an automatically selected arterial input function.⁶ DKI was acquired at the second time point using 30 directions with b-value=1000 s/mm², 2000 s/mm² and 10 b-value=0 s/mm² images (3x3x3 mm³) using simultaneous multislice image acquisition.⁷ Mean kurtosis (MK), axial kurtosis (AK), radial kurtosis (RK), mean diffusivity (MD), axial diffusivity (AD), radial diffusivity (RD) and fractional anisotropy (FA) maps were calculated using the Diffusion Kurtosis Estimator.⁸ For delineating region of interests (ROI), acute DWI, DKI, MRP and the follow-up infarct (FU) from the second MRI’s FLAIR images were co-registered to one another (MNI autoreg⁹), and to the MNI 152 1mm atlas. Abnormal perfusion was defined as tissue with Tmax values greater than 6 seconds. DWI and DKI were compared in the following regions: Core (abnormal acute DWI and FU), Growth (normal acute DWI, abnormal FU), and Salvaged (normal acute DWI, abnormal acute perfusion, normal FU). Contralateral normal tissue (CNL) was defined as tissue with no apparent abnormalities on acute and follow-up imaging. DWI and DKI metrics in the Core were correlated against time-to-MRI (nonparametric Spearman’s correlation analysis). To minimize errors due to poor co-registration, only patients >1 cm³ Growth or Saved ROIs were included for analysis. Each ROI was further segmented into gray matter (GM), white matter (WM) and cerebral spinal fluid. Mean acute DTI (MD, AD, RD, FA) and DKI values (MK, AK, RK) in the four ROIs: Core, Growth, Saved and CNL as a whole and in GM and WM areas were evaluated using one-way ANOVA with repeated measures followed by a post-hoc Tukey HSD test.

Patient characteristics were: mean±SD age 66.2±10.4 years, 72.2% (N=13) males, median [IQR] admission NIH Stroke Scale (NIHSS) 6 [2.75-11], time-to-acute MRI 6.2±2.1 h, time-to-F/u MRI 3.0 ± 1.3 days, 33% (N=6) received thrombolysis, acute DWI lesion 3.5 [0.6-20.4] cm³, acute Tmax lesion 10.5 [0-73.1] cm³, FU lesion 13.4 [1.8-56.3] cm³ and 90 day modified Rankin Scale 1 [1-2.25]. Figure 1 shows examples of the maps used for ROI placement, and example DTI and DKI maps. Increased kurtosis is evident, and most conspicuous in AK maps, in which normal GM and WM values are similar. Correlation between time-to-MRI and diffusion metrics were significant only for AD (ρ=-0.54, P=0.022), MK (ρ=0.37, P=0.049), and AK (ρ=0.55, P=0.017). Nine patients exhibited salvaged tissue. Significant differences were found between the ROIs for both GM (Figure 2) and WM (Figure 3). For GM, diffusivity metrics were significantly lower in all ROIs compared to CNL and were significantly with respect to one other. FA was lower in the Core than in Saved tissue for both GM and WM. In WM, Growth ROIs had greater FA values than Core, but smaller values than Saved and CNL. MK and AK values in the Core and Growth ROIs were significantly higher than both CNL and Saved ROI values for GM. For WM, this was true only for AK.

The significant inverse correlation between time-to-MRI and axial diffusivity changes in core tissue suggests that ischemia-induced cellular swelling may increase tortuosity of water diffusion paths, imposing direction-dependent restrictions upon diffusion. Coupled with changes in diffusivity, DKI may provide complementary insight into neuronal injury.² However, future studies are needed in the hyperacute stage to fully understand the role of DTI and DKI in identifying salvageable tissue.