Diffusion kurtosis imaging (DKI) is a useful tool to examine tissue microstructures based on measurements acquired at multiple b-values^1,2. On the other hand, recent studies suggested that diffusion MRI measurements acquired at multiple diffusion times can provide additional information on microstructural properties^3,4. Combining the two approaches may lead to more comprehensive understanding of microstructural organization, however, the relationships between the two approaches are not well understood. In this study, we measured diffusion kurtosis at varying diffusion times using pulsed and oscillating gradients, and attempted to characterize the diffusion time dependence of DKI in the normal and hypoxia-ischemia injured mouse brains.In vivo diffusion MRI were performed on a horizontal 11.7 Tesla scanner with a 72 mm volume transmitter and a 15 mm planar surface coil. Normal adult C57BL/6 mice (n = 7) were scanned using a pulsed gradient spin-echo (PGSE) sequence with δ/Δ = 4/20 ms and a cosine-trapezoid⁵ oscillating gradient spin-echo (OGSE) sequence with an oscillating frequency of 100Hz (equivalent diffusion time = 2.5 ms). The sequences were calibrated using a gel phantom. Images were acquired with a four-segment EPI readout, TE/TR = 55/3000 ms, NA = 4, 30 diffusion directions, b = 1000 and 2000 s/mm², in-plane resolution of 0.17 x 0.17 mm², and slice-thickness of 0.8 mm. Neonatal C57BL/6 mice (n = 5) at postnatal day 10 were subjected to unilateral hypoxia-ischemia (HI) using the Vannucci model⁶ and imaged at 24 hrs after injury with similar imaging parameters as the adult mice except in-plane resolution = 0.2 x 0.2 mm², slice-thickness = 1 mm, and NA = 2. Follow-up scans were performed at 3, 10, and 17 days after the injury with T2-weighted and DTI scans. The mean diffusivity (MD), mean kurtosis (MK), axial and radial kurtosis (K_// and K⊥) were obtained using the Diffusional Kurtosis Estimator software⁷.

In the normal adult mouse brain, MK_OGSE measured at 2.5 ms diffusion time was significantly lower than that MK_PGSE measured at 20 ms diffusion time (Fig. 1). The differences between PGSE and OGSE MKs (ΔMK= MK_PGSE - MK_OGSE, Fig. 1B) were greater in the cortex and hippocampus (77.8±3.5% and 41.2±6.3%, respectively) than in the corpus callosum and cerebral peduncle (33.7±3.4% and 23.8±4.4%, respectively) (Fig. 2A). In these white matter structures, the differences between PGSE and OGSE K_// were less than the differences in K⊥ (Fig. 2B-C).

In neonatal mouse brains that had severe edema (with hyper-intense T2 and reduced MD) at 24hrs after HI injury, both PGSE and OGSE MK were drastically elevated in the edema region (Fig. 3A). The ipsilateral cortical MK increased to more than three folds of the contralateral side (Table 1). Both PGSE and OGSE MK maps showed marked heterogeneity within the edema region. In addition, ΔMK was higher in the edema region than the contralateral side (Fig. 3B). Compared to other edema region, the dorsal hippocampus (indicated by the red arrows) had relatively normal PGSE/OGSE MK and ΔMK values (Table1), and this region was mostly spared at 17 days after injury without developing into cyst (Fig. 3C).

The reductions in MK when the diffusion time decreased from 20 ms to 2.5 ms agree with previous tissue diffusion modelling results at relatively short diffusion times⁸. It is likely that water diffusion becomes less restricted as diffusion time decreases, resulting in reduced non-Gaussianity. The change in diffusion kurtosis with diffusion time may reflect the spatial scale of microstructural barriers that cause the non-Gaussianity. For example, the greater ΔMK values in the cortex than the corpus callosum suggest a larger amount of water diffusion in the cortex turns less non-Gaussian as the diffusion distance shortens from 10 to 3 μm, given the diffusion time from 20 to 2.5 ms.

The investigation of diffusion kurtosis at varying diffusion time may offer additional information about the microstructural changes under pathological conditions. Recent studies showed the DKI has higher sensitivity than DWI in detecting brain injury^9,10,11, and the heterogeneity in kurtosis measurements may reflect the heterogeneity of injury. In the HI injured mouse brain, heterogeneity was observed in the PGSE and OGSE MK as well ΔMK images (Fig. 3), implying the involvement of microstructural changes at multiple spatial scales. The heterogeneity in the hippocampal region is worth investigating, and it may correlate with the long-term outcomes.