Structural Neuroimaging Sequences

2.8.1 Magnetic resonance imaging apparatus. Structural MRI, including DTI data were obtained using a Philips Achieva 3.0 T MRI scanner (Philips Medical Systems, Netherlands) using Nova Dual gradients, a phased array head coil, and sensitivity encoding (SENSE) with an under-sampling factor of 2. Quadratic shim gradients were used to correct for magnetic field inhomogeneities within the brain.

2.8.2 T1- and T2*-weighted magnetic resonance imaging sequences. High- resolution T1-weighted (MPRAGE) images were acquired with the following acquisition parameters: repetition time (TR) = 9.6 ms, echo time (TE) = 4.5 ms, 150 slices, slice thickness = 1.2 mm, flip angle = 8°, in-plane resolution = 0.94×0.94 mm, matrix size = 208x208 mm. Gradient-echo T2*-weighted images were obtained using the following sequence and parameters: echoplanar imaging with whole-brain coverage, TR = 2,000 ms, TE = 30 ms, 31 ascending slices, slice thickness = 3.25 mm, gap = 0.75 mm, flip angle = 90°, matrix size = 112×87 mm, field of view (FOV) = 280×220×123 mm.

2.8.3 Diffusion tensor imaging sequence. Diffusion-weighted volumes with gradients applied in 64 non-collinear directions were collected (16 directions per each of a total of four DTI runs). The following parameters were used: 73 contiguous slices, slice thickness = 2 mm, FOV = 224 mm, matrix = 128 x 128 (voxel size = 1.75 x 1.75 x 2 mm), b value = 1000 s/mm2, and four images obtained with no diffusion weighting (b = 0 s/mm2), one per each DTI run.

2.8.4 Preprocessing of diffusion tensor imaging data: A summary. Pre-processing

of the diffusion data was carried out using DTIStudio (Jiang, van Zijl, Kim, Pearlson, & Mori, 2006), a freely available resource program, and various image processing tools from the Oxford

Centre for the Functional Magnetic Resonance Imaging of the Brain’s (FMRIB) Software Library, FSL (Smith et al., 2004; Woolrich et al., 2009). The four pairs of PAR and REC files from the Philips Achieva scanner, each containing 16-direction DTI data, DTI sequence-specific gradient files, and associated bvals and bvecs were used as the starting point. All four DTI runs were first concatenated and the diffusion weighted images registered to the b = 0 s/mm2 images by affine transformations using the FMRIB’s Linear Image Registration Tool (FLIRT; Jenkinson & Smith, 2001; Jenkinson, Bannister, Brady, & Smith, 2002). This corrected the data for motion and eddy-current distortions due to the gradient coils that can compromise image quality by introducing stretches and shears. Brain extraction was then performed on the b = 0 image using the Brain Extraction Tool (BET; Smith, 2002) Version 2.1 to remove the skull and generate an image containing brain tissue only as well as a mask image. Due to individual differences in the relationship between the skull and brain tissue, the extraction factor used varied between individual participants. Each image was checked after the extraction, and where too much or too little had been extracted, as was often the case, the factor was adjusted accordingly until an accurate extraction was achieved (smaller values = larger brain outline).

Finally, dtifit from the FSL’s Diffusion Toolbox (FDT; Behrens et al., 2003) was used to perform the three-dimensional fitting of the tensor model on the data. The image input into dtifit was the data image derived as a result of the previous preprocessing steps. The mask image applied was that of the skull-extracted brain. In addition, the bvecs and bvals were input. Voxelwise fractional anisotropy (FA) and mean diffusivity (MD) maps were then generated, as well as images to quantify the three eigenvalues (λ1, λ2, and λ3) that represent the magnitude of diffusivity along each of the principal axes of the diffusion tensor (see Figure 2-1).

A series of 3D images describing various properties of the diffusion tensor thus resulted, including:

• dti_V1 – the first eigenvector • dti_V2 – the second eigenvector • dti_V3 – the third eigenvector

• dti_L1 – the first eigenvalue (λ1), i. e. parallel or ‘axial diffusivity’ • dti_L2 – the second eigenvalue (λ2)

• dti_MD – mean diffusivity

The trace (λ1 + λ2 + + λ3) of the diffusion tensor (D), is an invariant (i.e. orientation- independent) index of overall diffusion in a voxel or region. The mean diffusivity (MD) is given by trace(D) ÷ 3.

• dti_FA – fractional anisotropy

The following combination of the tensor eigenvalues λ1, λ2 and λ3 determines the fractional anisotropy FA of a voxel/region:

Axial (Dax) and radial (Drad) diffusivity images were also derived from these FDT outputs: Dax = dti_L1 (λ1)

Drad = dti_L2 (λ2) + dti_L3 (λ3) ÷ 2

Figure 2-1. Anisotropic and isotropic diffusion. The probability function for displacement of water molecules can be

depicted as an ellipsoid when diffusion is anisotropic as opposed to when diffusion is isotropic. The eigenvalues λ1, λ2, and λ3 represent the magnitude of diffusion along the three principal directions and the eigenvectors V1, V2, and V3 the length of each axis. Reprinted from Radiology December 2000, 217, Wiegell, M. R., Larsson, H. B. W., & Wedeen, V. J. Fiber crossing in human brain depicted with diffusion tensor MR imaging, 897-903, Copyright (2011), with permission from the Radiological Society of North America.

Figure 2-2 shows the first eigenvector (V1), corresponding to the orientation of the principal axis of the diffusion tensor, overlaid on the FA map of a healthy control participant.

Figure 2-2. Fractional anisotropy (FA) modulated by V1. A single axial slice (MNI Z = 17) of a healthy control

participant’s FA map with V1 (overlaid) displayed as red lines. On the right, a section of the corpus callosum is shown magnified; in this highly anisotropic white matter the principal diffusion direction (indexed by V1) is aligned with the longest axis of the white matter tract. MNI = Montréal Neurological Institute.

A more detailed description of the DTI data preprocessing steps is provided in Appendix C.