Voxel-based diffusion tensor imaging in patients with mesial temporal lobe epilepsy and hippocampal sclerosis

Abstract

Background

Mesial temporal lobe epilepsy (mTLE) with hippocampus sclerosis (HS) is an important cause for focal epilepsy. In this study, we explored the integrity of connecting networks using diffusion tensor imaging (DTI) and two whole-brain voxel-based methods: statistical parametric mapping (SPM) and tract-based spatial statistics (TBSS).

Methods

Thirty-three consecutive patients with mTLE and HS undergoing presurgical evaluation were scanned at 3 T, a DTI data set was acquired and parametric maps of fractional anisotropy (FA) and mean diffusivity (MD) were calculated. Twenty-one patients had left hippocampal sclerosis (LHS) and 12 patients had right HS (RHS). These groups were compared to 37 normal control subjects using both SPM5 and TBSS.

Results

The ipsilateral temporal lobe showed widespread FA reduction in both groups. The limbic system was clearly abnormal in the LHS group, also involving the arcuate fasciculus. In RHS, changes were more restricted but also showed involvement of the contralateral temporal and inferior frontal lobe. Increased MD was found in the ipsilateral hippocampus by SPM that was only marginally detected by TBSS. In white matter regions, however, TBSS was more sensitive to changes than SPM.

Conclusion

DTI detects extensive changes in mTLE with HS. The affected networks were principally in the ipsilateral temporal lobe and the limbic system but also the arcuate fasciculus. SPM and TBSS gave complementary information with higher sensitivity to FA changes using TBSS.

Introduction

Mesial temporal lobe epilepsy (mTLE) is the most common type of refractory focal epilepsy, and hippocampus sclerosis (HS) is the most frequent underlying pathology (Margerison and Corsellis, 1966). Resection of the sclerotic hippocampus is associated with seizure freedom or significant reduction of seizure frequency in the majority of patients (Berkovic et al., 1995, Helmstaedter et al., 2003). The hippocampus is part of the limbic system, comprising hippocampus, fornix, mammillary bodies, anterior thalamus, cingulum and the parahippocampus and has a key role in memory with a material specific lateralization for verbal memory in the dominant (usually left) and non-verbal memory in the non-dominant side (usually right) (Frisk and Milner, 1990, Powell et al., 2005).

Diffusion tensor imaging (DTI) is an MRI technique that can provide information about white matter fiber orientation and integrity in vivo. In white matter, diffusion perpendicular to the main fiber direction is more hindered by cell membranes and myelin layers than is diffusion along the main axis and is therefore anisotropic. To quantify this phenomenon, different indices have been devised that are rotationally invariant: the most commonly used is fractional anisotropy (FA) (Le Bihan et al., 2001, Pierpaoli and Basser, 1996).

Mean diffusivity (MD) reflects the total magnitude of diffusion in the corresponding voxel (Le Bihan et al., 2001). Studies using regions of interest (ROI) have shown changes in the hippocampus in unilateral mTLE with HS. The usual pattern is reduced FA and increased MD in the involved side but some contralateral changes may also occur (Assaf et al., 2003, Kimiwada et al., 2006, Salmenpera et al., 2006a, Wehner et al., 2007). DTI changes are not confined to the ipsilateral hippocampus alone but can also be found elsewhere (Arfanakis et al., 2002, Gross et al., 2006, Thivard et al., 2005).

The limitation of an ROI-based approach is that it is restricted to predefined regions and is subject to operator bias. To overcome this and allow study of the whole brain, voxel-based techniques have been developed, the most commonly used being statistical parametric mapping (SPM) (Ashburner and Friston, 2000). We have already shown the usefulness of voxel-based DTI analysis in MR negative patients and those with acquired lesions or malformations of cortical development (Eriksson et al., 2001, Rugg-Gunn et al., 2001). A methodological issue with the SPM approach is the spatial normalization that is required to make anatomically different subjects comparable in each individual voxel. Even with non-linear warps, it is usually not possible to achieve a perfect match; therefore, smoothing is necessary to account for some residual error and make the data conform more closely to the underlying model of Gaussian fields. Recently, a novel processing technique has been published that – instead of trying to match each and every voxel in different subjects – projects DTI data on a common pseudo anatomical skeleton and therefore does not need smoothing (Smith et al., 2006). This tract-based spatial statistics (TBSS) is available as part of the FSL software package.

Since most functions in the human brain are not carried out in a single anatomical structure but rather occur in systems, it is appropriate to analyze which connective networks show diffusion abnormalities in mTLE. We used both SPM and TBSS to investigate the changes in diffusion throughout the brain without prior constraints in patients with mTLE and HS. We also discuss methodological differences between the two voxel-based approaches.