White matter pathology following mild traumatic brain injury (mTBI) is complex and unlikely to be characterized by a single neuroimaging metric. DTI has been used to probe white matter microstructure, but is known to be non-specific. We show that the orientation dispersion index from the NODDI model may be more sensitive to white matter damage following mTBI in a mouse closed-skull impact model. Combined with quantitative susceptibility mapping, our data suggest minimal overt myelin loss, but progressive white matter injury, which may include myelin disruption, up to 6 weeks post-mTBI.
Mild traumatic brain injury (mTBI) can lead to highly variable cognitive and behavioural symptoms, and there is no way of predicting who will experience persistent symptoms. The brain’s white matter is particularly vulnerable to primary and secondary injury processes following mTBI, including axon damage and changes to myelin, but this microstructural damage is not detectable with conventional imaging techniques.1 There is a need to better detect and describe white matter damage following mTBI with neuroimaging in order to better predict outcome and evaluate emerging treatments.
Diffusion tensor imaging (DTI) is the most common technique used to probe white matter microstructure, but the metrics derived from the tensor model (fractional anisotropy (FA), mean, axial, and radial diffusivities (MD, AD, RD)) are inherently non-specific. Recent studies have suggested that more complex geometric models of diffusion, including the Neurite Orientation Dispersion and Density Imaging (NODDI) model2, may be more sensitive and specific to different aspects of white matter pathology in mTBI.3,4 Furthermore, MRI sequences that are specific to other properties of myelin besides diffusion, including quantitative susceptibility mapping (QSM), may help elucidate the role of myelin changes following mTBI.
The objective of this study is to assess white matter changes over time following mTBI using multi-shell diffusion weighted imaging (msDWI) and QSM in a mouse model of mTBI in order to examine the sensitivity and specificity of these techniques to mTBI pathology.
Mild TBI was induced in 8-week-old male C57BL/6J mice using a closed-skull impact model with the impact centred on the right parietal bone. Sham mice underwent the same surgical procedure without the impact. Beginning 3 days or 6 weeks following the injury, the mice underwent 5 days of behavioural testing to assess memory, anxiety and depression-like behaviours. Following the final behavioural test, the mice were perfusion-fixed for ex vivo MRI (n = 13/group/time point).
Brains were imaged within the skull using a msDWI protocol with FSE readout (30 directions at b = 7300, 20 directions at b = 5070, 15 directions at b = 2080, and 5 b = 0 images; 100 µm isotropic resolution; δ/Δ = 9 ms/17 ms; TR/first TE/echo spacing/ETL = 400 ms/38 ms/8 ms/6; scan time ~ 22 hrs) and with a spoiled gradient echo sequence (56 µm isotropic resolution; TE/TR/alpha=18 ms/100 ms/70°; scan time ~ 11 hrs) from which quantitative susceptibility maps were calculated using the phase images.5
The diffusion data were modelled with the tensor model using FSL (FMRIB, Oxford UK) to compute standard DTI metrics and with the NODDI model to obtain the orientation dispersion index (ODI), and the intracellular, extracellular, and isotropic (CSF) volume fractions2. The diffusion and susceptibility data were analyzed separately. In each case, the reference images were registered together to create a consensus average using an automated linear and non-linear registration algorithm. The transformations from these registrations were then applied to the calculated metrics maps. Statistical comparisons were made between the TBI and sham groups at each white matter voxel and in each structure of a segmented atlas (22 white matter structures per hemisphere)6 using a linear model. Multiple comparisons were corrected for using the false discovery rate (FDR).