OUP user menu

Diffusion tensor imaging in patients with epilepsy and malformations of cortical development

S. H. Eriksson, F. J. Rugg-Gunn, M. R. Symms, G. J. Barker, J. S. Duncan
DOI: http://dx.doi.org/10.1093/brain/124.3.617 617-626 First published online: 1 March 2001

Summary

Malformations of cortical development (MCD) are a common cause of epilepsy. Studies of structural MRIs and PET data in patients with MCD have found widespread changes outside the visually identified lesions. The aim of this study was to investigate diffusion changes interictally in patients with MCD and to test the hypothesis that MCD would be associated with widespread abnormalities of diffusion. We used diffusion tensor imaging (DTI) and statistical parametric mapping (SPM) to compare objectively tissue organization in 22 patients with partial seizures and MCD, with 30 control subjects. Whole-brain DTI was acquired using echo planar imaging. Rotationally invariant anisotropy and diffusivity maps were calculated and, after normalization to Talairach space, each patient was compared with the group of control subjects using SPM. Areas of reduced anisotropy were found in 17 patients and areas of increased diffusivity in 10. Two patients had areas of increased anisotropy. There were no patients with reduced diffusivity. Areas of increased diffusivity were in general more extensive than areas of reduced anisotropy. Changes in tissue beyond the MCD, that appeared normal on conventional MRI, were found in six patients for anisotropy and nine patients for diffusivity. Both measurements showed widespread changes in tissue beyond the MCD, i.e. adding information to conventional MRI. Increased or abnormally located grey matter or pathological white matter with abnormal myelination or ectopic neurones could cause reduced anisotropy. Increased diffusivity could be caused by a defect of neurogenesis or cell loss resulting in increased extracellular space. The widespread nature of abnormalities should be considered if surgical treatment is contemplated.

  • diffusion tensor imaging
  • malformations of cortical development
  • epilepsy
  • magnetic resonance imaging
  • BHT = band heterotopia
  • DTI = diffusion tensor imaging
  • DWI = diffusion-weighted imaging
  • EPI = echoplanar imaging
  • MCD = malformation of cortical development
  • ROI = region of interest
  • SCH = subcortical heterotopia
  • SEH = subependymal heterotopia
  • SPM = statistical parametric mapping

Introduction

Diffusion is a random process resulting from the thermal translational motion of molecules. The diffusion displacement distances are comparable with cellular dimensions, raising the possibility that the measurement of water diffusion might provide a means of exploring cellular integrity and pathology. Diffusion imaging incorporates pulsed magnetic field gradients into a standard MRI sequence resulting in images which are sensitive to the small displacements of water molecules. The diffusivity (a measurement of the magnitude of this diffusional motion) and anisotropy (a measurement of the directionality of the motion) can be calculated (Pierpaoli and Basser, 1996; Pierpaoli et al., 1996). In cerebral tissue, cell membranes restrict diffusion (Hajnal et al., 1991). Tissues with more hindered diffusion in one direction than another are said to be anisotropic (e.g. white matter) whereas diffusion is the same in all directions in isotropic tissue (e.g. CSF). The parallel orientation of fibres in tracts is an important determinant of anisotropic diffusion in the brain. Water molecules can diffuse relatively freely parallel to, but only to a limited extent perpendicular to, tracts. In densely packed unmyelinated fibres, the many axonal membranes may restrict diffusion and, in myelinated fibres, the multiple myelin laminae could cause the restriction (Beaulieu and Allen, 1994; Nomura et al., 1994; Wimberger et al., 1995). Thus diffusion tensor imaging (DTI) gives microstructural data about tissues and provides further information on, for example, white matter tracts that cannot be obtained using conventional T1- or T2-weighted images. Pathological processes that change the microstructural environment, such as neuronal swelling or shrinkage, increased or decreased extracellular space and loss of tissue organization, result in altered diffusion and/or anisotropy (Sevick et al., 1992; Anderson et al., 1996).

It has been shown that diffusion-weighted imaging (DWI) (using diffusion gradients in only three directions) underestimates the magnitude of the anisotropy in tracts that are oblique to the diffusion gradients (Basser et al., 1994; Pierpaoli and Basser, 1996; Pierpaoli et al., 1996). DTI is a development of DWI in which diffusion gradients are applied in at least six non-collinear directions. Using DTI, the total diffusion tensor can be calculated and the resulting diffusion parameters are insensitive to the subject's position and the fibre tract alignment within the scanner.

In humans, DWI has been used mainly for the early diagnosis of acute ischaemic stroke (Warach et al., 1995; Lutsep et al., 1997). Diffusion changes have also been seen in multiple sclerosis (Droogan et al., 1999) and brain tumours (Tien et al., 1994). In epilepsy, reduced diffusivity has been shown in animal models of status epilepticus (Zhong et al., 1993, 1995; Nakasu et al., 1995; Ebisu et al., 1996; Wang et al., 1996), most probably due to aberrant ionic homeostasis causing shifts in intra- and extracellular water, or cytoplasmic dysfunction (Lux et al., 1986; Duong et al., 1998). In patients with partial status epilepticus, reduced diffusivity has been demonstrated using DWI, in cerebral areas corresponding to seizure semiology, EEG (Wieshmann et al., 1997) and electrocorticography (Diehl et al., 1999). In interictal DWI studies of patients with chronic epilepsy and hippocampal sclerosis, both increased diffusivity (Hugg et al., 1999; Wieshmann et al., 1999a) and reduced anisotropy (Wieshmann et al., 1999a) were found in the sclerotic hippocampi, suggesting a loss of structural organization and expansion of the extracellular space.

Malformations of cortical development (MCD) are associated commonly with epilepsy, and are often identified with high-resolution MRI (Palmini et al., 1991; Brodtkorb et al., 1992; Raymond et al., 1995; Chan et al., 1998). The malformations are the result of defects of neuronal proliferation, migration or organization during brain development. As the development of the grey and white matter are closely linked, malformations can affect both the cortex and the underlying white matter (Barth, 1987). Widespread changes outside the MCD visible on conventional MRI have been found using analyses of the regional distribution of grey and white matter on T1-weighted images (Sisodiya et al., 1995; Woermann et al., 1999a) and [11C]flumazenil-PET (Richardson et al., 1996).

The aim of the study was to investigate interictal changes of diffusivity and anisotropy in patients with MCD using DTI and to test the hypothesis that MCD would be associated with widespread abnormalities of anisotropy and diffusivity. We used statistical parametric mapping (SPM) to make voxel-by-voxel comparisons between each patient and a group of 30 control subjects (Friston et al., 1995a, b).

Material and methods

Subjects

This study was approved by the ethical committee of the National Hospital for Neurology and Neurosurgery. Thirty healthy volunteers with no history of neurological disorder and normal standard MRI formed the control group (20 women and 10 men, median age 30 years, range 20–50 years). We studied 22 patients with partial seizures and MCD (13 women and seven men, median age 31 years, range 21–51). The patients were recruited from the clinics of the National Society for Epilepsy and the National Hospital for Neurology and Neurosurgery and had MCD previously demonstrated by MRI. Clinical details and EEG findings are presented in Table 1. Only one patient (Patient 3) had a complex partial seizure less than 24 h before the DTI scan. All subjects gave informed consent to participate in the study.

View this table:
Table 1

Clinical characteristics, EEG, MRI and DTI findings in 22 patients with MCD

PatientAge (years)GenderSeizure typesEEG featuresMRI findingsDTI findings: decreases in anisotropy (max P level)Increases in anisotropy (max P level)Increases in diffusivity (max P level)
abn. = abnormalities; ant. = anterior; BHT = band heterotopia; bil. = bilateral; F = female; front. = frontal; gen. = general; GM = grey matter; L. = left; M = male; MCD = malformation of cortical development; med. = medial; myocl. = myoclonic; NA = normal appearing; occ. = occipital; par. = parietal; post. = posterior; R. = right; S/CPS = simple/complex partial seizure; SEH = subependymal heterotopia; SGTC = secondary generalized tonic–clonic seizure; temp. = temporal; WM = white matter.
127FSGTCBil abn. max LBil frontopar. gyral abn with thick cortexBil frontopar. in MCD and front. bil. in NA GM and WM (P < 0.0001)L. front. med. of MCD (P = 0.003)Bil frontopar. in MCD. Bil front. and occ. in NA GM, WM and cerebellum (P < 0.0001)
224FSPS, CPS, tonicBil abn. max LBil gen. gyral abn. with thick frontopar. cortexBil frontopar. and R. front. in MCD (P < 0.0001)NoneBil frontopar. in MCD. Bil front, temp-occ. in NA GM, WM and cerebellum (P < 0.0001)
322MSPS, CPSBil abn. max LBil frontopar. gyral abn. with thick cortexBil frontopar. in MCD (P < 0.0001)NoneNone
424MCPS, SGTCBil abn. max temp.Bil frontopar. gyral abn. with thick cortexBil frontopar. in MCD (P = 0.03)NoneNone
534FSPS, CPS, SGTCNo definite abn.L. frontopar. gyral abn. with thick cortexL. frontopar. in MCD (P < 0.0001)NoneL. NA cerebellum (P = 0.001)
639FSPS, CPSBitemp. abn.L. front. and R. frontopar. schizencephalyL. front. and R. frontopar. in MCD (P < 0.0001)NoneNone
720MSGTCBitemp. abn. max LGen gyral abn. with thick cortexNoneNoneBil widespread occ. to front. in MCD and NA WM (P = 0.03)
836FSPS, CPS, SGTC myocl. jerksBitemp. abn. max LGen gyral abn. with thick cortexBil in MCD (P < 0.0001)NoneR. parieto-occ. and L. occ. in MCD Bil temp. in NA WM (P = 0.001)
950MCPS, SGTCR. abn. max temp.R. front. and bil. par. and occ. gyral abn. with thick cortexBil in MCD. L. front. in NA WM (P < 0.0001)NoneR. widespread temp. to front. in MCD and underlying NA WM. L. front. to par. WM (P < 0.0001)
1044MCPS, SGTCL. hemisphere abn.L. front. and par. gyral abn. with thick cortexL. frontopar. in MCD. R front. in NA GM (P = 0.01)NoneL. frontopar. in MCD and underlying NA WM, bil cerebellum (P < 0.0001)
1125FSPS, myocl jerksR. hemispere abn. max ant.R. par. and occ. gyral abn. with thick cortexL. pericallosal in NA GM and WM (P = 0.02)NoneR. frontopar. in MCD and underlying WM (P < 0.0001)
1231FSGTCBil abn.Agenesis of corpus callosum, gen. gyral abn. and SEHBil in MCD and area of corpus callosum (P < 0.0001)NoneNone
1351MSGTCNo definite abn.R. par. nodular heterotopiaR. par. in MCD (P < 0.0001)NoneNone
1423MCPS, SGTCR. several foci, max post. temp. and parieto-occ.R. frontopar. and R. occ nodular heterotopiaR. frontopar. in MCD (P <0.0001)R. occ. in WM adjacent to MCD and L. temp. in GM (P = 0.002)R. frontopar. and occ. in MCD and adj WM. Bil cerebellum and NA GM and WM L occ-temp. (P < 0.0001)
1531FSPS, CPS, SGTCBil abnBil BHTNA L. post. periventricular GM and WM (P < 0.0001)NoneNone
1621MCPS, SGTCGen abn. ictal bil. centralBil BHTNoneNoneNone
1750FCPS, SGTCNo focal abnBil BHTNoneNoneNone
1827MSPS, CPS, SGTCBil abn. max L tempBil BHTNoneNoneNone
1926FSPS, CPSBil abn. max LBil SEHBil in MCD (P < 0.0001)NoneR. occ-temp. NA GM and WM (P < 0.0001)
2036FSPS, SGTCNo focal abnBil SEHL. in MCD (P = 0.02)NoneNone
2132FCPSL. temp. abnL. post. SEH and L. front. gyral abn. with thick cortexL. post. and L. front. in MCD. L. occ. in NA GM and WM (P = 0.002)NoneNone
2232FCPS, SGTCNo focal abnR. post. SEHNoneNoneNone

Conventional MRI scanning protocol

All control subjects and patients had conventional MRI on a 1.5 T Horizon Echospeed scanner (GE, Milwaukee, Wis., USA) [T1-weighted IRp-SPGR volumetric acquisition, contiguous 5 mm oblique coronal T2-weighted, proton density and fast FLAIR (fluid attenuated inversion recovery) images]. These were reviewed by experienced neuroradiologists who detected no abnormalities in the control subjects.

DTI scanning protocol

Scans were performed on a 1.5 T Horizon Echospeed scanner (GE, Milwaukee, USA). Single-shot CSF-suppressed diffusion-weighted echoplanar imaging (EPI) was used [TR/TE/TI (repetition time/echo time/inversion time) = 5000/78/1788 ms], acquisition matrix 96 × 96, reconstruction matrix 128 × 128, FOV (field of view) 24 cm, slice thickness 5 mm covering the whole brain (Barker et al., 1997). Pulsed unipolar diffusion gradients were used for diffusion sensitization (δ = 28 ms, Δ = 35 ms). Two b-values, degree of diffusion weighting (b = 0 and bmax = 703 s/mm2), were applied in seven non-collinear directions: x, y, z, xy, xz, yz, and xyz (to allow rotationally invariant parameters to be calculated) at 13 slice positions. Two interleaved series with nine repeats each were acquired, resulting in 1872 images (there was no interslice gap). Repeat scans were averaged after magnitude reconstruction to increase the signal-to-noise ratio while retaining the low motion sensitivity of the single-shot acquisition. Diffusion scanning time was 19 min. Total scanning time including diffusion, localizer and high-resolution EPI anatomical scan was 25 min. Images were transferred to a separate workstation (Sun Microsystems, Palo Alto, Calif., USA) for post-processing.

Maps of fractional anisotropy and mean diffusivity (rotationally invariant scalar indices of diffusion) were calculated from the diffusion-weighted images according to the method proposed by Pierpaoli and Basser (Pierpaoli and Basser, 1996; Pierpaoli et al., 1996) using software written in-house. The fractional anisotropy map represents a value of the anisotropy index, with 0 representing an isotropic medium where there is no directionality of the diffusion and 1.0 representing maximum anisotropy. The magnitude of the diffusion in each voxel is represented by the mean diffusivity measured in ×10–6 mm2/s.

SPM analyses

To allow voxel-by-voxel statistical comparison to be made, all images were spatially normalized to a standard template using SPM96 (Wellcome Department of Cognitive Neurology, Institute of Neurology, London, UK) (Friston et al., 1995a, b). Since the contrast of anisotropy and diffusivity maps differs from the T1- and T2-weighted templates provided by SPM96 and there are geometric distortions associated with the EPI acquisition, a new template was created. This was created by normalizing a control subject's image with no diffusion weighting (the b = 0 image) to Talairach space using 12 linear degrees of freedom and a 4 × 4 × 5 non-linear warp. The b = 0 images of all control subjects and patients were normalized to the template using linear transformations with 12 degrees of freedom (translation, rotation, zoom and shear) which provided a robust normalization. Non-linear normalization was not used in the control subjects and patients to avoid inappropriate elimination of focal individual differences that would reduce sensitivity of the analysis. The origin for normalization was set to the anterior commissure. The anisotropy and diffusivity maps were then normalized using the parameters determined from the normalization of the b = 0 image. The normalized anisotropy maps were smoothed with an 8-mm and the diffusivity maps with a 10-mm isotropic Gaussian kernel to improve the signal-to-noise ratio, to increase the validity of statistical inference and to improve normalization (Friston et al., 1995b). The signal intensity threshold was set at 0.5 to reduce the number of low signal-to-noise voxels included in the analysis. The signal intensity threshold is a proportion of the whole-brain image mean intensity, and only voxels exceeding this intensity threshold (50% of the mean in our study) were included in the analyses. In all patients and control subjects, this procedure excluded noise areas outside the brain or in the ventricles.

Two contrasts were used to detect whether each voxel had a higher or lower anisotropy or diffusivity in each individual patient compared with the group of control subjects. Significant increases or decreases were detected at an individual voxel threshold of P < 0.001. To correct for multiple comparisons, the resulting foci were characterized in terms of spatial extent (corrected P value P < 0.05) (Friston et al., 1994). Each individual patient was compared with the control group. Similarly, each control subject was compared with the remainder of the control group using equivalent smoothing, signal intensity thresholds and significance levels.

Region-of-interest analyses

Since SPM will only show that there is a significant focal difference between an individual subject and the control group, quantitative region-of-interest (ROI) analyses were made in three patients whose appearance on conventional MRI were considered to be typical for each aetiology, to indicate the magnitude of the differences seen. For decreases of anisotropy, ROI were analysed in a patient with bilateral gyral abnormalities and thickened cortex (Patient 1, Fig. 1) and a patient with focal neuronal heterotopia (Patient 14, Fig. 2). For increased anisotropy, ROI parameters were determined in a patient with focal neuronal heterotopia (Patient 14, Fig. 2). For increased diffusivity, ROI were analysed in a patient with bilateral gyral abnormalities and thickened cortex (Patient 1, Fig. 3) and a patient with unilateral gyral abnormalities (Patient 11). The areas detected by SPM as deviating significantly from normal in terms of voxel intensity and cluster extent (P < 0.05) were outlined automatically and thereafter superimposed on the normalized anisotropy and diffusivity maps using DispImage 4.7 software (Plummer, 1992). The mean value in the ROI on the patient's map was compared with the mean value in the same ROI from all 30 of the control subjects' maps, and the difference expressed as a percentage of the control mean.

Fig. 1

Patient 1: bilateral frontoparietal gyral abnormalities with thickened cortex. (A) Normalized axial anisotropy maps at the same slice localization for the averaged 30 control subjects. (B) The patient. Note that the difference in the signal-to-noise ratio between the two maps is due to averaging of the 30 control subjects. (C) Regions of significantly decreased anisotropy identified with SPM are superimposed (blue) on the patient's normalized anisotropy map. (D) The equivalent slice of the patient's T1-weighted image. The regions of decreased anisotropy not only coincide with the localization of the gyral abnormalities with thickened cortex, but are also found in the normal-appearing occipital lobes. Note that right on the images is the patient's right.

Fig. 2

Patient 14: right frontoparietal and occipital nodular heterotopia. (A) normalized axial anisotropy maps at the same slice localization for the averaged 30 control subjects. (B) The patient. Note that the difference in the signal-to-noise ratio between the two maps is due to averaging of the 30 control subjects. (C) Regions of significantly decreased (blue) and increased anisotropy (red) are superimposed on the patient's normalized anisotropy map. A region of reduced anisotropy is found in the frontoparietal heterotopia. (D) The equivalent slice of the patient's T1-weighted image. The region of increased anisotropy is in normal-appearing white matter adjacent to the occipital heterotopia. Note that right on the images is the patient's right.

Fig. 3

Patient 1: bilateral frontoparietal gyral abnormalities with thickened cortex. (A) Normalized axial diffusivity maps at the same slice localization for the averaged 30 control subjects. (B) The patient. Note that the difference in the signal-to-noise ratio between the two maps is due to averaging of the 30 control subjects. (C) Regions of significantly increased diffusivity identified with SPM are superimposed on the patient's normalized diffusivity map (red). (D) The equivalent slice on the patient's T1-weighted image. The regions of increased diffusivity not only coincide with the localization of the gyral abnormalities, but are also found in normal-appearing white matter in parietal and frontal lobes. Note that right on the images is the patient's right.

Results

Control subjects

Comparisons of the anisotropy and diffusivity maps of each normal subject with the remaining 29 control subjects revealed decreased anisotropy in one individual and increased diffusivity in two individuals at the statistical threshold of P < 0.001 with a correction for multiple comparisons at P < 0.05. Since 60 tests were made (30 examinations for each contrast), up to three areas would be expected by chance for each diffusion index at a significance level of P < 0.05. Group analyses of the control subjects comparing left versus right handedness and male versus female showed no significant differences.

Anisotropy

Decreased anisotropy

In 17 of the 22 patients with MCD, SPM detected areas of decreased anisotropy (Table 1). In 15 of these patients, the changes corresponded to all or part of the MCD. In six patients, changes were found outside the MCD in tissue that appeared normal on conventional MRI.

Nine of the 11 patients (Patients 1–11) with gyral abnormalities had areas of significantly reduced anisotropy that corresponded to all or part of the MCD. In four of the 11, changes were found in areas beyond the margins of the evident MCD, in areas that appeared normal on T1- and T2-weighted images (Fig. 1). Reduced anisotropy was also seen in the patients with agenesis of the corpus callosum (Patient 12) and subcortical heterotopia (SCH) (Patients 13 and 14, Fig. 2). None of the patients with band heterotopia (BHT) (Patients 15–18) had any significant anisotropy changes detected by SPM in the areas of the MCD. One patient (Patient 15), however, had a periventricular area of decreased anisotropy outside the evident MCD. In the patients with subependymal heterotopia (SEH), SPM detected areas of reduced anisotropy that corresponded to all or part of the SEH in three patients (19–21). In one patient (Patient 21), reduced anisotropy was also shown in the frontal gyral abnormality and in normal-appearing grey and white matter.

Increased anisotropy

In two patients (Patients 1 and 14), SPM detected areas of increased anisotropy. Two of the areas were in white matter adjacent to MCD (Fig. 2). In one patient (Patient 14), a further area of increased anisotropy was found in the left temporal lobe in grey matter that appeared normal on conventional MRI.

ROI analyses

Quantitative ROI analyses of decreased anisotropy were made in Patients 1 and 14, whose appearances on conventional MRI were typical of gyral abnormality with thickened cortex and nodular heterotopia, respectively. In Patient 1 (Fig. 1), the analyses were made in the normal-appearing left occipital white matter and right frontoparietal area with gyral abnormality. The mean anisotropy value was 0.35 in the left occipital region, which was 60% of the mean of control values of 0.58. The mean anisotropy in the the right frontoparietal region was 0.30. This was 51% of the mean anisotropy in the corresponding areas in the control subjects of 0.59. In the right frontoparietal heterotopia in Patient 14 (Fig. 2), the mean anisotropy was 0.43, which was 61% of the mean anisotropy value (0.71) in the same region in the control subjects.

ROI analyses were also made in one of the patients with areas of significantly increased anisotropy, Patient 14 (Fig. 2). In this area of normal-appearing occipital lobe, the anisotropy value was 0.75 in the patient and 0.38 in the control subjects (197%).

Diffusivity

Increased diffusivity

In 10 of the 22 patients with MCD, SPM detected areas of increased diffusivity (Table 1). In eight of these, the changes corresponded to all or part of the MCD. In nine patients, changes were found outside the MCD in tissue that appeared normal on conventional MRI. In seven of the 10 patients, the diffusivity changes were more extensive than the anisotropy changes.

Seven of the 11 patients (Patients 1–11) with gyral abnormalities had areas of increased diffusivity corresponding to all or part of the MCD. In all of these and one additional patient, changes were found outside the MCD in grey and/or white matter that appeared normal on conventional MRI (Fig. 3). One of the patients with nodular SCH (Patient 14) had areas of increased diffusivity both within and in areas beyond the MCD. There were no diffusivity changes in the other patient with nodular SCH (Patient 13), the patient with agenesis of the corpus callosum (Patient 12) or the patients with BHT (Patients 15–18). One (Patient 19) of the patients with SEH (Patients 19–22) had an area of increased diffusivity outside the MCD in normal-appearing grey and white matter.

Decreased diffusivity

There were no patients with regions of significantly reduced diffusivity.

ROI analyses

Quantitative ROI analyses were made in two patients whose appearances on conventional MRI were typical of gyral abnormality with thickened cortex. In Patient 1, the mean diffusivity for an area of normal-appearing white matter in the left occipital lobe (Fig. 3) was 811 × 10–6 mm2/s, compared with a mean of 712 × 10–6 mm2/s for the control group in the same area (114%). In the normal-appearing white matter in the left frontal lobe in Patient 9, the average diffusivity was 873 × 10–6 mm2/s. This was 119% of the average diffusivity in the corresponding area in the control group, 731 × 10–6 mm2/s.

Discussion

Using voxel-by-voxel comparisons of DTI, 15 patients had areas of reduced anisotropy and eight had areas of increased diffusivity within the MCD. Reduced anisotropy was also found in tissue beyond the margins of the evident MCD in six patients, in areas that appeared normal on conventional T1- and T2-weighted images. Nine of the patients had increased diffusivity in grey and/or white matter that appeared normal on MRI. In two patients, there were areas of increased anisotropy. There were no patients with reduced diffusivity. There were no diffusion changes in the patients with BHT; apart from this, there was no apparent relationship between the nature of changes and type of MCD.

In all patients, anisotropy images added information about the structure of the white matter tracts that was not seen on conventional MRI. In addition, in six patients, areas of decreased anisotropy and, in nine patients, areas of increased diffusivity were detected outside the margins of the lesions seen on T1- and T2-weighted images. This suggests that despite the low sensitivity in detecting MCD, DTI detected subtle changes not visualized with optimal conventional MRI.

Methodological aspects and limitations

Ultra-fast EPI has enabled us to perform whole-brain DTI within a reasonable time frame. The rotationally invariant diffusion parameters calculated using this technique are insensitive to variations in intersubject positioning within the scanner, allowing meaningful comparison between individuals (Basser et al., 1994; Pierpaoli and Basser, 1996). Since DTI is sensitive to molecular movement of water, it is also sensitive to motion and pulse artefacts. To reduce the effect of motion artefacts, we used a fast single-shot MRI sequence (EPI) for the acquisition. The resolution of this whole-brain DTI was 2.5 × 2.5 × 5.0 mm, which limits the size of the changes that can be seen. This voxel size was a compromise between imaging time and an adequate signal-to-noise ratio. To increase the signal-to-noise ratio further, nine averages were acquired at every slice position.

The voxel-by-voxel approach used by SPM is a more statistically rigorous and unbiased method than ROI analyses based on a priori knowledge. SPM previously has been applied successfully to structural MRI (Woermann et al., 1999a, b) and PET data (Richardson et al., 1996, 1997). In addition, an ROI-based study of DTI in structural cerebral lesions failed to identify abnormalities in mean diffusivity in 30% of patients despite anisotropy abnormalities in all, and did not investigate normal-appearing cerebral tissue (Wieshmann et al., 1999b). Voxel-by-voxel comparisons are only possible if all brains are in a common space, making normalization necessary to reduce the overall variability in size and shape between brains. Visual assessment showed that normalization worked satisfactorily in all subjects, even in patients with severely disordered brains. This is supported by the small number of changes seen in the analyses of the control subjects and in areas outside the brain in patients. The statistical threshold used was deliberately rigorous to avoid false-positive findings. Inevitably, this may have reduced the sensitivity in the patient group. Several factors further limited the sensitivity of the methods employed. The human cortical ribbon is 2–5 mm thick and, with the DTI voxels being of comparable size, any abnormalities had to be of a sufficient size to be significant with the stringent statistical approach used. The normalization scheme based on the Talairach atlas is most successful with central structures; therefore, these regions will be matched most accurately to the other subjects after normalization. The most peripheral cerebral structures, for example cortical ribbon and subcortical white matter, will correspond less well, resulting in a wider range of anisotropic values in the control subjects. Statistical comparison of these areas between controls and patients may therefore fail to identify subtle differences.

Biological and clinical implications

The areas of reduced anisotropy detected within MCD were in what appeared to be grey matter and/or grey–white matter interface on high-resolution MRI. The decreased anisotropy in these areas of heterotopic grey matter is likely to be caused by the comparisons of anisotropy in the abnormally located grey matter in the patients with anisotropy in white matter in control subjects.

More interestingly, anisotropy and diffusivity changes were found outside the MCD in normal-appearing tissue. Changes beyond the margins of the visually detected MCD were found in nine of the 10 patients with increased diffusivity, and in six of the 17 patients with decreased anisotropy, concurring with previous studies of structural MRI (Sisodiya et al., 1995; Woermann et al., 1999a) and PET data (Richardson et al., 1996, 1997) of patients with epilepsy and MCD. These findings suggest that MCD is often more extensive than the visible lesion, with widespread subtle malformation.

Several histopathological epilepsy surgery and necropsy series have found microdysgenesis and an increased number of neurones in white matter in patients with epilepsy compared with control subjects (Meencke, 1983; Hardiman et al., 1988; Kasper et al., 1999). These changes may involve several lobes (Eriksson et al., 1999). An increased number of neuronal cell bodies in the white matter (as in microdysgenesis) could disrupt the white matter tracts and cause a reduction in anisotropy. Other studies have found a relative decrease in white matter in patients with MCD, that might be the result of an increased number of neurones projecting thinner axons (Sisodiya et al., 1995), which may have altered arborization (Mitchison, 1991) and altered anisotropy. Poor myelination of white matter may also occur with MCD (Marchal et al., 1989), and this might contribute to reduced anisotropy. Histopathological correlations of our DTI findings are not available at present. Since patients with MCD do not commonly proceed to surgical treatment, because they are unlikely to do well (Bruton, 1988; Engel, 1993), post-mortem studies will be necessary.

Increases in diffusivity generally were more extensive than decreases in anisotropy, in both MCD and normal-appearing grey and white matter. The widespread increase in diffusivity suggests areas of reduced cell density and increased extracellular space due to failure of neurogenesis or later cell loss. In animal studies of status epilepticus, reduced diffusivity has been seen in the acute phase with spontaneous normalization after 3 days (Nakasu et al., 1995). The animals were only followed for 1 week and it is unclear if the values would continue to rise, as after a stroke (Warach et al., 1995; Lutsep et al., 1997). However, in the subsequent histopathological examinations, pyknotic neurones and vacuolated neuropil were seen in the areas with previously reduced diffusivity, suggesting neuronal damage sustained during the prolonged seizure and which would be expected to result in increased extracellular space and increased diffusivity. Repetitive seizures in humans may also result in neurone loss and gliosis (Vinters et al., 1993), and both reduced anisotropy and increased diffusivity might be caused by frequent seizures.

In two patients, there was an increase in anisotropy in normal-appearing white matter adjacent to MCD. This most probably reflects the comparison of white matter tracts, that have been displaced by the MCD, with grey matter in controls or of fibres in white matter tracts being more densely packed, resulting in an increased anisotropy (Beaulieu and Allen, 1994).

SPM did not detect significant changes in anisotropy or diffusivity in all MCD. Anisotropy is usually highest in the major white matter tracts and lower in the tissue close to the cortex where fibres are crossing or fanning out (Peled et al., 1998). MCD in areas where anisotropy is naturally low might not result in significant reduction in anisotropy compared with control subjects. Anisotropy was, however, more sensitive in detecting the MCD than was diffusivity. Our findings are concordant with previous ROI-based analyses of a group of three patients with MCD that all had anisotropy changes but only one had diffusivity changes (Wieshmann et al., 1999b). Features common to all MCD are loss of tissue organization and abnormalities of neuronal structure. This may cause reduced anisotropy. Despite the disorganization of the tissue, the cellular density is preserved in many of the MCD. The diffusivity restriction by cell membranes and diffusivity value might therefore still be similar to that in normal tissue. This suggests that anisotropy and diffusivity are diffusion entities that give complementary information. The pattern of changes is different to that seen in acquired lesions causing epilepsy in which diffusivity changes were more evident than abnormalities of anisotropy (Rugg-Gunn et al., 2001).

The finding of widespread changes in anisotropy and diffusivity could be of potential clinical importance if surgical treatment is considered in patients with epilepsy and apparently focal MCD and in the investigation of occult abnormalities in cryptogenic epilepsy.

Several studies have shown reduced diffusivity during partial status epilepticus correlated with the cortical areas involved (Wieshmann et al., 1997; Diehl et al., 1999; Lansberg et al., 1999). In the current study, all scans were interictal and no reductions of diffusivity were found. In selected patients, ictal DTI may be a useful non-invasive tool for localization of the epileptogenic focus. It is not known, however, if there will be an ictal reduction in diffusivity in patients who have an interictal increase in diffusivity. In view of this and the heterogeneity of the results, any ictal diffusivity maps would need to be interpreted in the light of interictal data.

Acknowledgments

We wish to thank Drs Brian Kendall and John Stevens for expert neuroradiological review of images and assistance, and Drs Friedrich Woermann and Udo Wieshmann for helpful discussions. This study was supported by Action Research. S.H.E. is supported by the Swedish Epilepsy Society, Glaxo Wellcome AB and Holmia Försäkringar AB. F.J.R-G. is supported by Action Research. M.R.S. is supported by the Brain Research Trust. G.J.B. is supported by the Multiple Sclerosis Society of Great Britain and Northern Ireland. J.S.D. is supported by the National Society for Epilepsy (NSE). The Glaxo Wellcome Scanner and the MRI Unit are supported by the NSE.

References

View Abstract