Magnetic Resonance Imaging

MRI studies were performed on a 3T GE Signa Excite HDx scanner (General Electric, Waukesha, Milwaukee, WI, USA). Standard imaging gradients with a maximum strength of 40mT m-1and slew rate 150T m-1s-1 were used. All data were acquired using a body coil for transmission, and eight channel phased array coil for reception.

6.5.3.1

Structural Imaging

Standard clinical scans were performed with the sequences listed in Table 6.4.

6.5.3.2

Diffusion Tensor Imaging

6.5.3.2.1 Acquisition

DTI data were acquired using a cardiac-triggered single-shot spin-echo planar imaging (EPI) sequence (Wheeler-Kingshott et al. 2002) with echo time (TE) of 73 ms. Sets of 60 contiguous 2.4mm thick axial slices were obtained covering the whole brain, with diffusion sensitising gradients applied in each of 52 non-collinear directions [b value of 1200 mm2s-1 (δ = 21 ms, ∆ = 29 ms, using full gradient strength of 40 mT m-1)] along with six non-diffusion weighted scans. The gradient directions were calculated and ordered as described elsewhere (Cook et al. 2007). The field of view was 24x24cm, and the acquisition matrix size was 96x96, zero filled to 128x128 during reconstruction giving a reconstructed voxel size of

1.875x1.875x2.4mm. The DTI acquisition time for a total of 3480 image slices was approximately 25 minutes (depending on subject heart rate).

Accompanying field maps were acquired from December 2011 until the end of the project. A 3D fast spoiled gradient echo sequence (FSPGR) sequence with 64 contiguous 2.4mm thick axial slices was performed covering the same region as the DTI and ensuring no table movement between the two acquisitions. Two slices were discarded at each end and the sequence was performed twice using a repetition time (TR) of 7.6ms and a TE of 5.46ms and 3.00ms respectively. The field of view was 24x24cm and the acquisition matrix size was 128x128, zero filled to 256x256 giving a reconstructed voxel size of 0.9375x0.9375x2.4mm. Real and imaginary images were acquired and used to calculate phase maps.

Sequence Plane Acquisition

time (m:ss) Slice thickness / spacing (mm) Number of slices Matrix size Field of view (cm)

T1 FLAIR Sagittal obl 1:28 5 / 2 17 320x224 24x24

FSPGR 3D Coronal obl 7:30 1.1 / 0 170 256x256 18x24

T2 FLAIR Coronal obl 2:56 5 / 0 32 224x256 18x24

PD/T2 FRFSE

Coronal obl 2:01 5 / 0 32 256x256 18x24

FGRE T2* Coronal obl 2:12 5 / 0 32 192x192 18x24

FSE T2 Axial obl 1:23 5 / 0 30 256x512 18x24

PROPELLER

(selected patients)

Coronal obl 3:20 2 / 0 32 416x416 22x22

Table 6.4 - Epilepsy protocol for structural scans on 3T GE Signa Excite HDx scanner

Key: FGRE = fast gradient echo, FLAIR = fluid-attenuated inversion recovery, FRFSE = fast recovery fast spin echo,

FSE = fast spin echo, FSPGR = fast spoiled gradient recalled, PD = proton density, PROPELLER = periodically rotated overlapping parallel lines with enhanced reconstruction, SE = spin echo, obl = oblique

6.5.3.2.2 Preprocessing

The scans were transferred to a Linux-based Sun Ultra workstation in DICOM format and converted to a single multivolume Analyze 7.5 format file using locally written software. Eddy current correction of the DTI data was performed using the eddy_correct tool in FSL (Smith et al. 2004).

A multi-tensor model was fitted to the eddy corrected diffusion data using the Camino toolkit (version 2 release 767) (Cook et al. 2006). Voxels in which a single tensor fitted the data poorly were identified using a spherical-harmonic voxel-classification algorithm (Alexander et al. 2002). In such voxels a two tensor model was fitted, with the principal diffusion directions of the two diffusion tensors providing estimates of the orientations of the crossing fibres. In all other voxels, a single tensor model was fitted.

6.5.3.2.3 Tractography of the Optic Radiation

Tractography was carried out using the Probabilistic Index of Connectivity (PICo) algorithm (Parker et al. 2003) as implemented in Camino and extended to deal with multiple fibres (Parker & Alexander 2003, Cook et al. 2004). Seed, way and exclusion masks were defined using MRIcro (http://www.psychology.nottingham.ac.uk) based on previous work (Yogarajah et al. 2009).

Fractional anisotropy and principal diffusion direction maps were used to identify the lateral geniculate nucleus (LGN) by selecting the axial slice at the level of the transition from the posterior limb of the internal capsule to the cerebral peduncle. Voxels antero-lateral to the LGN across the base of Meyer's loop, with principle eigenvectors orientated in an antero-medial to postero-lateral orientation, were identified and used to define a seed point in a coronal plane (Figure 6.1a-c). Contiguous voxels, with principal directions in an anterior-posterior direction, were also selected in order to cover the entire coronal cross-section of Meyer's loop, using a standardised seed point volume of 15 voxels (127 mm3).

Tracking from the seed was performed using 50000 Monte Carlo iterations, an angular threshold of 180° and a fractional anisotropy threshold of 0.1, in order to ensure that the paths detected would not erroneously enter areas of cerebrospinal fluid, and yet had sufficient angular flexibility to allow tracking of Meyer's loop.

In order to restrict the pathway to anatomically valid tracts, a way point was defined in the lateral wall of the occipital horn of the lateral ventricle at the posterior extent of the corpus callosum (Figure 6.1d). Two exclusion masks were applied. Firstly, a midline exclusion mask and then a coronal exclusion mask to remove artefactual connections to adjacent white matter tracts, such as the fronto-occipital fasciculus, anterior commissure and uncinate fasciculus. An objective, iterative process was performed to determine the optimum location for this mask whereby the exclusion mask was moved posteriorly until it began to coincide with Meyer’s loop, identified by a visible thinning of the estimated trajectory of the optic radiation, typically associated with a reduction in tract volume greater than 10% (Yogarajah et al., 2009). A connectivity distribution was generated from each voxel in the seed region and combined into an overall connectivity map representing the maximum observed connection probability to each voxel within the brain from all the voxels within the seed region. For display purposes, the connectivity distributions were thresholded at 5%, representing a compromise between retaining anatomically valid tracts and removing obviously artefactual connections.

The above approach to tractography was slightly modified for surgical planning outside the temporal lobe in Chapter 7 and two different methods were compared in Chapter 8.

Figure 6.1 - Seed and way point for optic radiation tractography

The seed point was across Meyer’s loop, red directions showing the high curvature of the loop, C sagittale (lateral wall of the lateral ventricle) (D

6.5.3.3

Functional MRI

Subjects also underwent five

 Working memory (visuospatial n  Working memory (verbal n  Language (verbal fluency)  Language (verb generation)

 Episodic memory (memory encoding)

Seed and way point for optic radiation tractography

The seed point was across Meyer’s loop, red (A – axial slice, B – close up of axial slice with principle diffusion directions showing the high curvature of the loop, C – coronal slice) and the way point

all of the lateral ventricle) (D). Displayed on the FA map.

Functional MRI

s also underwent five fMRI paradigms:

Working memory (visuospatial n-back) Working memory (verbal n-back) Language (verbal fluency) Language (verb generation)

Episodic memory (memory encoding)

close up of axial slice with principle diffusion the way point, green, was in the stratum

For the working memory fMRI task, gradient-echo planar T2*-weighted images were obtained with 50 non-contiguous 2.4mm thick oblique axial slices (0.1mm gap) covering the whole brain. The field of view was 24x24cm and the acquisition matrix size was 64x64 giving in-plane resolution of 3.75x3.75mm. TE was 25ms, TR was 2.5s and a SENSE factor of 2 was used.

The working memory network was identified using a modified version of the ‘n-back’ task (Callicott et al. 1999, Kumari et al. 2003) in which subjects were required to monitor the location of dots (presentation time 440ms, inter-stimulus interval 1500 ms) within a diamond shaped box on the screen at a given delay with the original occurrence (0-, 1- or 2-back). There were three 30 second active conditions in total (0-, 1-, and 2-back) presented to subjects five times in pseudorandom order, controlling for any order effect. In total, 15 stimuli were presented in each 30 second active block. Each active condition started with a 15 second resting baseline (the word “Rest” appeared on the screen during this period). Subjects were required to move the joystick corresponding to the correct location of the current (0-back) or previously presented (1-back = previous presentation; 2-back = previous presentation but one) stimulus. On-line accuracy data were determined by joystick movement on every trial with output stating either a correct response, wrong response or no response. The percentage of correct 2 dot-back trials was used as a measure of performance.

The other functional MRI paradigms were not used in this thesis.