MRI studies

MRI data were acquired using a 1.5 T GE Signa Echospeed MRI scanner with maximum gradient strength of 33 mTm ⁻¹ . All individuals were scanned at baseline, 3, 6 and at 12 months.

The MRI protocol was the follow-up study of Jenkins et al. (2011) , where baseline data were reported. For the present study, we processed the images corresponding to 3-, 6-, and 12-month follow-up time points. The protocol included:

The following scans were carried out: (i) coronal-oblique proton density fast spin-echo sequence (repetition time = 2300 ms, echo time = 68 ms, two excitations, echo train length = 8, matrix size = 512 × 384, field of view = 24 × 18 cm, 16 contiguous 3-mm slices); (ii) coronal-oblique fluid-attenuated inversion-recovery (FLAIR) imaging (repetition time = 2500 ms, echo time = 12.7 ms, inversion time = 995 ms, six excitations, echo train length = 6, matrix size = 512 × 384, field of view = 24 × 18 cm, 16 contiguous 3-mm slices); and (iii) post-triple-dose Gd-enhanced coronal-oblique fat-saturated T ₁ -weighted spin-echo was acquired (repetition time = 600 ms, echo time = 20 ms, one excitation, matrix size = 256 × 192, field of view = 24 × 18 cm, 16 contiguous 3-mm slices).

Optic nerve images were processed to obtain: (i) lesion length (only in patients, at all time points), by an experienced (and blinded) neuroradiologist (K.M.); (ii) optic nerve cross-sectional area (all individuals, at all time points), from five contiguous slices anterior from the orbital apex ( Hickman et al. , 2001 ), using a semiautomated contouring technique ( Plummer, 1992 ), by a blinded observer (T.J.), as a marker of optic nerve axonal loss ( Trip et al. , 2006 ); and (iii) Gd-enhanced lesion length (only in patients and only at baseline)

Visual functional MRI task-dependent scans were acquired in four scanning sessions, as previously described ( Jenkins et al. , 2010 a ).

For diffusion-weighted imaging scans of the optic radiations and occipital lobes, we used an optimized single-shot, cardiac-gated, diffusion-weighted echo-planar imaging sequence, as previously described ( Jenkins et al. , 2010 a , b , 2011 ). Diffusion-weighting gradients were applied along 61 distributed directions ( Cook et al. , 2007 ), with b = 1200 s/mm ² , which was optimized for white matter. Seven interleaved non-diffusion-weighted b0 scans were also acquired. One additional b0 volume was acquired, covering the whole brain to assist co-registration of partial brain diffusion data to whole-brain functional MRI data (see below), which was necessary for tractography. Head motion- and eddy current-induced distortions were corrected and the diffusion tensor was then calculated on a pixel-by-pixel basis, using FSL tools ( http://www.fmrib.ox.ac.uk/fsl ).

Axial oblique, proton-density, dual echo, fast spin echo scans of the whole brain were carried out as previously described ( Jenkins et al. , 2010 a ).

Processing of optic radiation images was as follows. Diffusion-weighted images: the optic radiations were reconstructed using the FSL probabilistic tractography algorithm ( http://www.fmrib.ox.ac.uk/fsl/fdt/fdt_probtrackx.html ). First level functional MRI contrast images for all subjects were combined and two spherical regions of interest of 3.5-mm radius were created centred on the global maximal coordinates for each lateral geniculate nucleus (MNI coordinates: right, 22 −24 −4; left, −22 −26 −4). Regions of interest underwent normalization to each subject’s native space, co-registration to each subject’s partial brain diffusion data, via the whole brain b0 intermediate step, and binarization. These regions of interest were then moved in each subject’s native diffusion space to sit 8–10 voxels laterally within the apex of Meyer’s loops. Visual confirmation that they were placed correctly was performed in each case and probabilistic tractography was performed from these seed points ( Jenkins et al. , 2010 a ). The reason why functional MRI seeds were averaged at the group level was that signal at the individual level was considered too variable, and tractography results would have been unreliable. The mean FA, AD and RD within the tractography-derived tract were obtained for each side, in each subject, at each time point. Because optic nerve fibres approximately hemi-decussate at the optic chiasm, DTI metrics for left and right optic radiation sides were averaged, after checking there were no significant differences between right and left hemispheres (data not shown).

Proton density images were used to obtain the whole brain and optic radiation lesion load.

Visual functional MRI task-dependent scans were used to define the seed points needed for tractography of the optic radiations.