Advanced Diffusion Imaging in The Hippocampus of Rats with Mild Traumatic Brain Injury

The protocol has been approved by the Animal Ethics Committee at the University of Ghent (ECD 15/44Aanv), and all experiments were conducted in accordance with the guidelines of the European Commission (Directive 2010/63/EU).

1. Animal preparation and helmet attachment

1. Weigh a female Wistar H rat (± 250 g or 12 weeks of age) and anesthetize in a small induction chamber filled with a mixture of isoflurane (5%) and O₂ for at least 1 min.
2. Inject the rat with 0.05 mg/kg buprenorphine subcutaneously in the neck, return it to the home cage, and allow pre-emptive analgesia for at least 30 min to take an effect.
   NOTE: During the 30 min wait, the surgical site can be prepared.
3. Place a heating pad kept at 37 °C under the surgical field. Place the sterilized surgical instruments on the surgical field that was disinfected with 70% ethanol.
4. Place the rat back in the induction chamber and anesthetize the rat until it is non-responsive to a paw or tail pinch.
5. Place the rat on the surgical field and insert a catheter in the lateral tail vein. Next, shave the rat’s head, remove excess fur and disinfect the scalp and the rest of the surgical area with chlorohexidine. 
6. Inject 100 μL of 2% lidocaine locally in the scalp.
7. Make a midline incision using a scalpel size 11 to expose the skull, removing any excess membranes with small scissors. Retract the skin using an ocular speculum with a maximum spread of 1 cm.  Additionally, remove the periosteum by gently rubbing a sterile cotton bud across the skull until the periosteum is no longer present.
8. Put one drop of tissue glue on the skull and one on the sterilized metallic disc (diameter of 10 mm and 3 mm thickness), which acts as the helmet. Glue the disc approximately one-third before and two-thirds behind the bregma. Allow the glue to dry for 1 min.

2. Induction of traumatic brain injury (TBI)

1. Place the rat on the custom-made bed with a foam mattress of certain spring constant (see Table of Materials). Position the rat directly under a transparent plastic tube with a 450 g brass weight with the helmet as horizontal as possible. Detach the rat from anesthesia.
2. Pull the weight up to 1 m and release when ready. Ensure that a second experimenter is present to move the rat away from the plastic tube immediately after impact to prevent a second impact.
   NOTE: The sham injured rats receive the same experimental procedure (steps 1.1–2.7), except for step 2.2.
3. Re-attach the rat to the anesthesia and inject 1 mL of physiological solution (0.9% NaCl) via the catheter to reduce the hemodynamic shock.
   NOTE: It is possible that the rat briefly stops breathing because of the impact. Gently compress the thorax if the rat does not spontaneously breathe after 2 s to encourage the breathing reflex.
4. Remove the helmet by gently pulling it from the skull. Remove any remaining glue from the skull and skin and close the incision with surgical suture. Apply local analgesia gel using a sterile applicator tip. 
5. Place the rat on the bed of the CT scanner. Confirm the correct position using a scout scan. Adjust the field of view to enable imaging of the whole head within one bed position. Administer a general purpose, low dose CT scan to rule out skull fractures.
   NOTE: Skull fracture is a criterium for euthanasia.
6. Place the rat in a clean cage on a heating pad (37 °C). Monitor the time to regain consciousness. Once the rat is able to sit upright, the rat can be returned to the home cage.
7. Administer a second dose of 0.05 mg/kg buprenorphine one day following TBI induction.

3. Diffusion magnetic resonance imaging (MRI)

NOTE: Diffusion-weighted imaging is performed before and 1 day following trauma induction.

1. Anesthetize the rat in a small induction chamber filled with a mixture of isoflurane (5%) and O₂. When the rat is non-responsive to a paw or tail pinch reduce the anesthesia to 2% with a flow rate of 500 mL/min. Transfer the animal to the scanner bed in head-first prone position.
2. Position the rat in the head holder with the teeth bar and nose cone, delivering the anesthesia, and slide the head forwards until the center of the brain is at the level of the center of the quadrature volume MRI coil. Apply lubricating ointment to the eyes in small amounts to prevent any damage to the cornea. Fixate the head with a small piece of tape to avoid movement during scanning.
3. Place a pressure pad under the thorax of the rat to monitor respiration and cover the rat with a circulating warm water heating blanket and bubble wrap to keep the rat warm. Before the scanning, check the respiratory monitor to ensure that the signal is clear without noise and that the respiratory cycle is consistent. Relocate the pressure pad, if necessary.
   NOTE: The respiratory rate should be kept between 1 breath per 1,200–1,700 ms by adjusting the level of anesthesia between 1%–2%.
4. Slide the quadrature volume coil over the head. Adjust the tuning and matching capacitors of the coil to the proper frequency and impedance according to the instructions provided by the coil vendor. Advance the scanner bed into the scanner bore to start scanning.
5. Obtain a default three-plane scout scan (“tripilot”) to ensure correct positioning.
   1. Load the tripilot sequence into the Scan Control by clicking New Scan and selecting the tripilot sequence from the protocol list. Next, click the traffic light button to start the scan.
   2. When the scan is finished, load the scan in the image display and ensure that 1) the head is lying straight and 2) the brain is positioned in the center of the magnet and coil. Adjust the position of the head and/or the scanner bed, if necessary, and acquire a new tripilot scan.
6. Adjust the local magnetic field using an automated second-order shimming protocol: load the second-order shim protocol into the Scan Control as described in step 3.5.1. Next, click on the Acq tab | Current Adjustments | Method specific adjustment for the Local Field Homogeneity in the Spectrometer Control Tool window to start automated shimming.
7. Load a new T2 Rapid imaging with Refocused Echoes (RARE) sequence into the Scan control as described in step 3.5.1.
   1. Acquire T2 weighted images using the default settings, except for the following parameters:
   2. Open the Edit Scan tab and adjust the repetition time (TR) and echo time (TE) to 5,500 ms and 37 ms, respectively. Also, modify the field of view and matrix size to allow for a higher in-plane resolution of 109 μm x 109 μm (default resolution = 156 μm x 156 μm). Make sure that the slice thickness is 600 μm, number of slices is set to 45, and RARE factor is set to 8.
   3. Open the Geometry editor and place the slice package in the correct position including the bulbus of the brain and the cerebellum.
8. Load three new echo-planar diffusion-weighted spin-echo sequence (DtiEpi) from the B_DIFFUSION folder into the Scan Control protocol as described in step 3.5.1.
   NOTE: Using three different diffusion “shells”, the diffusion tensor imaging (DTI) model^4,16, diffusion kurtosis imaging (DKI) model¹⁷, and white matter tract integrity (WMTI) model¹⁸ can all be estimated. It is recommended to use at least three different b-values, with the highest b-value having a maximum of 3000 s/mm² with at least 15 evenly spaced directions per imaging shell¹⁷.
   1. Acquire diffusion weighted images (DWIs) using default settings, apart from the following settings:
   2. Open the Edit Scan tab and adjust the geometrical parameters under the Geometry tab. Adjust the field of view and matrix size to 105 x 105 to ensure a resolution of 333 μm x 333 μm.
   3. Set the slice orientation to axial and the number of slices to 25, resulting in a slice thickness of 500 μm and interslice distance of 600 μm. Amend the readout direction into left-right.
   4. Click the Contrast tab to adjust the echo time to 24 ms and repetition time to 6,250 ms.
   5. Set the bandwidth to 250,000 Hz and turn the fat suppression on. Adjust the number of averages to one.
   6. Click on the Investigación tab and change the number of averages (EPI segments) to 4.
   7. Click on the Diffusion tab within the research tab. Perform this step separately for each of the three diffusion shells.
      1. Adjust the number of diffusion directions to 32 for the first shell, 46 for the second shell, and 64 for the third shell.
      2. Adjust the gradient directions with custom gradient directions files.
      3. Change the number of B0 images to 5 for the first shell, 5 for the second shell, and 7 for the third shell.
      4. Adjust the b-value per direction to 800 s/mm² for the first shell, 1500 s/mm² for the second shell, and 2000 s/mm² for the third shell.
         NOTE: Adjusting the gradient directions with a custom gradient directions file can be done manually by setting Enter Diffusion Directions to yes or automatically by using the DTI_SET_DIRECTIONS macro.
   8. Open the Geometry editor and place the field of view between the bulbus and cerebellum containing only the cerebrum to reduce artifact and scan time. Position six saturation bands of 5 mm outside the brain to reduce artifacts by clicking on Saturation and sliding the bands in the preferred position using the scroll bars.
      NOTE: The bulbus and cerebellum can be identified based on anatomical landmarks and the three images of the tripilot scan.
9. Acquire the imported sequences by clicking on the traffic light symbol. Using the settings of the parameters described above, the acquisition time of the T2-RARE scan is 12 min, of the first DWI shell 15 min, of the second DWI shell 21 min and of the third shell 30 min. The total acquisition time is approximately 80 min (on a single receiver channel system).
10. At the completion of the scanning protocol, remove the animal from the scanner bed, and place the animal in a clean cage with a heating pad at 37 °C. Return the animal to the home cage when it regains consciousness.

4. Image processing

NOTE: In the following sections, the processing of the diffusion images is described in MRtrix3, ExploreDTI¹⁹ and Amide software²⁰ which are open access toolboxes. However, the preprocessing steps can be performed in other toolboxes (e.g., FSL, MedInria, DTIStudio).

1. Transfer the acquired data from the acquisition console by exporting the 2dseq file.
2. Convert the 2dseq files (raw DWI files) into the .mif format, which is the standard formatting of MRtrix3, to allow for further preprocessing steps in MRtrix3. Moreover, concatenate the three diffusion shells using the following commands in the shell:
   convert_bruker pdata/1/2dseq ratID_T2.mih (for the T2-weighted images)
   convert_bruker pdata/1/2dseq ratID_dwi1.mih (for the first diffusion shell)
   convert_bruker pdata/1/2dseq ratID_dwi2.mih (for the second diffusion shell)
   convert_bruker pdata/1/2dseq ratID_dwi3.mih (for the third diffusion shell)
   mrcat ratID_dwi1.mif ratID_dwi2.mif ratID_dwi3.mif ratID_dwi.mif
3. Perform noise correction and Gibbs ringing correction on the DWIs in MRtrix3^21,22. Also, convert the corrected DWI images and T2 image to NIFTI format using the following commands:
   dwidenoise ratID_dwi.mif ratID_dwi_denoised.mif
   mrdegibbs ratID_dwi_denoised.mif ratID_dwi_denoised_gr.mif
   mrconvert ratID_dwi_denoised_gr.mif ratID.nii
   mrconvert ratID_T2.mif ratID_T2.nii
4. Perform correction for EPI, motion and Eddy current distortions in ExploreDTI:
   1. Convert the NIFTI images to a .mat file by clicking Calculate DTI*.mat file | Convert raw data to DTI*.mat file. Change the diffusion tensor estimation to weighed linear and the b-value to NaN. Adjust the voxel size to 0.333 0.333 0.6, number of non-DWI images to 17, number of DWI images to 142, and matrix size to 105 105 25.
      NOTE: By setting the b-value to NaN, ExploreDTI will regard the dataset as a kurtosis data set.
   2. Click on the Settings tab to adjust the settings for EPI correction (this is turned off by default). Select SM/EC/EPI correction, also register to other data? and click on Yes, to do the EPI correction (non-rigid). Specify the suffix of the anatomical T2 image corresponding to the diffusion data set.
      NOTE: ExploreDTI corrects for EPI distortions using image registration between the undistorted anatomical image and the diffusion image.
   3. Click on the Plugins tab and select Correction for subject motion & EC/EPI distortions and select the preprocessed diffusion data file from step 2.3. Make sure that the T2 image is in the same folder and has the same base as the diffusion data file name (e.g., rat1.nii for the DWI and rat1_T2.nii for the anatomical image). This step will generate a “native” (*native.mat) and “transformed” file (*trafo.mat).
5. Calculate the DTI metrics for each rat by clicking Plugins and Export stuff to *.nii and selecting the parametric maps of the DTI model: fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD), and axial diffusivity (AD; denoted as “largest eigenvalue L1”).
6. Additionally, export the parametric maps for the kurtosis model (MK, AK, and RK) and WMTI model (AWF, AxEAD, RadEAD and TORT). Processing of the diffusion images will result in 12 parametric maps (Figure 1, Figure 2, Figure 3) that can be used for further microstructural analysis.
7. Create a mask file for the hippocampus of each rat using MRtrix3.
   1. Load the FA image of the rat in the MRtrix viewer by clicking Tool and ROI editor.
   2. Create a new ROI by clicking the “+” button and press Edit to draw the ROI on each slice that includes the hippocampus (Figure 4). To erase unwanted areas from the ROI drawn, press the right mouse button.
   3. When the drawing of the ROI is completed, save the mask image by clicking the Save button.
      NOTE: This mask file will be a binary NIFTI image file with voxels of value 1 containing the hippocampal tissue, and remaining voxels will have values of 0. To standardize the region of the hippocampus across rats, the parametric maps can be co-registered with a study specific template with predefined regions of interest delineated²³ or a rat brain atlas.
8. To extract the diffusion metrics of the hippocampus of the rat, use the created mask file of step 4.6 and open the Amide software.
   1. Open the parametric maps and mask image of the rat.
   2. To add the ROI of the mask file into Amide, select the mask file image, click Edit | Add ROI | 3D Isocontour and click on the ROI displayed in the mask image. Give the ROI a meaningful name and confirm that this volume should only contain voxels having a value of one.
   3. To calculate the mean values of the diffusion metrics in the hippocampus, click Tools | Calculate ROI Statistics and indicate the images and ROI that should be included. After clicking Execute, another screen will pop up with computed values that can be used for further statistical analysis. This file can either be saved or copied into a preferred data format (e.g., .xlsx or .csv file).

5. Statistical analysis

NOTE: In the following sections, we describe processing of the diffusion images in SPSS Statistics 24; however, the statistical analysis can be performed in other statistical toolboxes.

1. Load the data in the wide format in an SPSS *.sav file.
2. To test for statistical differences between the two groups for each timepoint (i.e., baseline or 1-day post-injury), click Analyze | Nonparametric tests | Legacy Dialogs | 2 Independent Samples Tests. Load the variables that need to be tested and specify the groups (i.e., TBI and sham groups). Indicate the Mann-Whitney U as the test type.
3. To test for statistical differences between the 2 time point within each group the data file needs to be split. Go to Go to Data, Split File and indicate Compare groups. Next, click on Analyze, Nonparametric tests, Legacy Dialogs, 2 Related Samples Tests, load the variables that need to be compared and indicate Wilcoxon as test type.
   NOTE: To correct for multiple comparisons, p-values are adjusted for each diffusion model using Bonferroni correction [i.e., p-value divided by the number of parameters compared (DTI 4, DKI 3, and WMTI 4)]. More specifically, p < 0.0125 is considered significant for the DTI and WMTI models, and p < 0.016 is considered significant for the DKI model.