Introduction

Structural MRI is a crucial tool routinely used for neuroanatomical presurgical evaluation of patients with refractory cryptogenic partial epilepsy (CPE) for the aetiological characterisation of the epileptogenic cortex. However, conventional MRI fails to identify cerebral lesions in approximately 25% of patients with CPE, and surgical treatment of MRI negative patients is frequently associated with a poor post-surgical outcome.1 Advanced MRI techniques provide the opportunity to identify subtle structural abnormalities that may underlie the epileptogenic cortex, thus increasing the likelihood of successful surgical remediation of intractable seizures.

Diffusion tensor imaging (DTI) is an MRI technique that can indirectly evaluate white matter (WM) integrity by measuring the spatial directionality of water diffusion. Some DTI studies have suggested that structural abnormalities in patients with CPE are not only restricted to easily identified circumscribed lesions. However, there are few studies that have addressed the utility of DTI in patients with CPE and normal conventional MRI. Most studies include heterogeneous samples of patients with CPE, with and without hippocampal sclerosis or circumscribed lesions, and results were inconsistent.2–6

In the present study, we sought to combine DTI with statistical parametric mapping (SPM) to determine fractional anisotropy (FA) abnormalities across the entire brain in patients with CPE for whom conventional MRI revealed no structural pathology.

Methods

Subjects

For DTI analysis, 12 patients (eight women, mean age of 39±10.4 years) with cryptogenic partial seizures and 67 healthy volunteers (24 women, median age 31.8±4.1 years) with no history of neurological disease were scanned. All epilepsy patients were recruited from our epilepsy monitoring unit. The conventional imaging sequences (T1, T2, FLAIR) of all examined patients did not show any abnormalities by visual inspection. All epilepsy patients underwent medical and neurological examination, interictal EEG and long term video EEG monitoring (at least 48 h). All patients had refractory partial seizures evolving to secondarily generalised seizures and gave written informed consent prior to examination. CPE was diagnosed according to clinical and EEG features following the International League Against Epilepsy classification system.7 Seizures were classified according to the semiological seizure classification.8 Detailed clinical characteristics of all patients are listed in table 1.

MRI data analysis

Image data were obtained on a 3.0 T system with a high resolution structural T1-weighted three-dimensional turbo field echo sequence (reconstructed after zero filling to 512×410×320 cubic voxels, edge length 0.5 mm), as well as T2-weighted and FLAIR imaging. For DTI we employed echo planar imaging with 20 diffusion directions (36 slices, thickness 3.6 mm, matrix 128×128, inplane resolution 1.8×1.8 mm).

After correction for eddy currents, the echo planar imaging images were spatially normalised to the Montreal Neurological Institute coordinate system following an optimised procedure.9 The diffusion tensor and FA field maps of all participants were calculated from spatially normalised images (inhouse software), normalised to an FA template image, also corresponding to the Montreal Neurological Institute coordinate, and smoothed by an 8 mm isotropic Gaussian kernel. Differences in FA values between the patient group and healthy controls were evaluated using SPM (SPM5, ANCOVA, modelling age as a covariate, p<0.001). To assess the magnitude of regional FA alterations between patients and controls, we also performed additional quantitative region of interest (ROI) analyses (see table 2). To define these subregions, we used an averaged and symmetrised (x axis) mask of all healthy controls with FA values >0.4.

Because of the reported local FA alterations of CPE patients in the literature, we expected the most asymmetrical changes in the temporal lobes. To assess the asymmetry between the mean temporal lobe FA values, an asymmetry index was defined as asymmetry index (AI)=(100×(FA left temporal lobe–FA right temporal lobe)/(FA left temporal lobe+FA right temporal lobe)).

Brain tissue volumes, normalised for subject head size, were calculated from the high resolution T1-weighted images, using the cross sectional version of the Structural Imaging Evaluation of Normalised Atrophy (SIENA) software (SIENAx).10

Results

Groupwise FA differences

Post hoc t tests of SPM–ANCOVA revealed significant FA reductions in WM areas of patients, covering widespread parts of the brain, most prominent in the frontal and temporal lobes, corpus callosum, corticospinal tracts and the brainstem. In ROI analysis, only FA values of the cerebellum, occipital lobe, parietal lobes and thalamus did not differ significantly from the control subjects (table 2).

Individual FA analysis

Comparisons of individual CPE patients with control subjects revealed diffuse FA alterations as well as distinct circumscribed FA ‘drop-outs’ in several WM regions (figure 1). The extratemporal WM abnormalities occurred mainly symmetrically while all CPE patients showed pronounced asymmetric FA reductions in the temporal lobes. In five patients, FA was particularly reduced in the right temporal lobe, and in seven patients primarily the left temporal lobe. Although only four patients had electroclinical evidence of temporal lobe epileptogenic focus, lateralisation of the temporal FA changes concurred with the affected hemisphere in all patients where available evidence allowed lateralisation of the presumed epileptogenic zone (table 2). Two patients without a lateralised focus (patient Nos 4 and 5) and one patient (patient No 3) with a bitemporal seizure origin showed FA reductions lateralised to the left temporal lobe.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Fractional anisotropy (FA) abnormalities observed in two patients with cryptogenic partial epilepsy relative to the cohort of control subjects (one sample t test, p<0.001; uncorrected; 20 contiguous voxels). The statistical map was superimposed on an averaged FA template of the control group. Examples of voxels with asymmetrical FA decrease. The presumed epileptogenic zone of patient No 1 and patient No 7 was localised in the right and left hemispheres, respectively. Lateralisation of the temporal FA changes concurred with the affected hemisphere in all patients with a defined electroclinical focus (nine patients). The circumscribed temporal (white arrows) and extratemporal (yellow arrows) FA reductions in the statistical maps were also clearly visible in corresponding colour coded FA maps (framed figures).

Asymmetry analysis

Group analysis of the AI showed significant asymmetrically increased mean FA values of the temporal lobes of patients compared with healthy controls (p=0.015). Additionally, individual AI values lateralised to the hemisphere of the epileptogenic zone in all nine patients (table 2) with a defined focus. In three patients without a defined electroclinical focus, there was an asymmetric decrease in left temporal FA values.

Global volumetric differences

There were no differences in brain volumes of CPE patients (grey matter, WM, relative and normalised brain volumes) compared with the group of healthy controls.

Clinical correlations

No significant correlation was found between the mean FA reduction in any ROI or the asymmetry indices and duration of habitual epilepsy, frequency of seizures, age of onset of epilepsy, presumed localisation of the epileptogenic zone (by EEG or seizure semiology), number of generalised tonic–clonic seizures or duration of anticonvulsant therapy.

Discussion

We have reported a symmetric widespread FA reduction in MRI negative patients with CPE relative to controls, which provides further evidence for architectural brain abnormalities outside the epileptogenic zone in these patients. In addition, we found significant reductions in FA in all patients, which occurred symmetrically in extratemporal regions. However, FA reduction in the temporal lobes was consistently identified ipsilateral to the seizure focus.

Our results are consistent with other neuroimaging studies that have indicated structural or functional abnormalities extending beyond the presumed epileptogenic zone.4 6 11–14 Some authors have postulated that widespread microstructural changes in epilepsy patients could be related to seizure activity and thus disease severity, resulting in extensive and reversible changes of neurons.11 We found no evidence of an association between FA alterations and duration or age of onset of habitual epilepsy, duration of the anticonvulsant medication, incidence of generalised tonic–clonic seizures or EEG localisation. Although our relatively small sample size does not allow us to make firm conclusions, this is at least consistent with studies showing no correlation with clinical factors and structural or metabolic abnormalities, suggesting that these extrafocal changes might be related to a more complex process of epileptogenesis.6 14

Our findings extend previous studies that identified FA changes in heterogeneous samples of patients with and without known structural abnormalities. Using DTI and individual statistical comparison, we identified FA alterations in temporal lobe regions consistently ipsilateral to the putative seizure focus in patients with no observable brain abnormality on conventional MRI, which suggests that focal FA abnormalities may underlie the epileptogenic region that belies visual inspection of T1-weighted MRI. Various studies showed numerous anatomical and functional connection bundles of the temporal lobe.11 14–16 The temporal FA changes of the affected hemisphere could be a sign of chronic remodelling of temporal structures due to repeated involvement in seizure originating from the epileptogenic focus, and the ipsilateral temporal WM in particular might be vulnerable to such impulses.17 The fact that the temporal lobe was predominantly affected may be a functional correlate of atypical language organisation in patients with partial epilepsy.18 Recently, a study revealed an association between these functional changes with temporal WM tract abnormalities in CPE patients.5 Such anatomical and functional temporal changes may be clinically important because it may interfere with various routes of seizure propagation.

Lateralisation of the affected hemisphere in patients with partial epilepsy, particularly cases without abnormal findings on high resolution MRI, remains a challenge for presurgical evaluation. An underlying MRI abnormality is a favourable prognostic indicator for seizure freedom after surgical treatment.1 However, in this study the seizure focus of the partial epilepsy was located in the temporal lobe in only four patients. Thus despite its sensitivity, DTI did not identify a direct clinical concordant abnormality in CPE patients with normal conventional MRI, and therefore this technique alone appears to have limited power in presurgical evaluation. Nevertheless, DTI might help to provide further information beyond what can be derived from standard epilepsy imaging protocols—that is, identifying the hemisphere enclosing the epileptogenic focus in cryptogenic epilepsy and particularly focusing on temporal changes might be helpful.

We suggest that WM FA analysis may represent an effective method of lateralising a neuroanatomical abnormality underlying the seizure focus during presurgical evaluation of CPE, in particular when supplemented with other neuroimaging techniques (MRS, fluorodeoxyglucose-positron emission tomography, quantitative T2 mapping). In order for this potential to be realised, patients need to be further investigated to correlate presurgical FA alterations with post-surgical outcome, as only then will a direct relationship between microstructural alterations and the epileptogenic zone be established.

In conclusion, DTI seemed to be much more sensitive than conventional MRI in demonstrating WM abnormalities in CPE. We found no evidence to suggest that FA alterations were not associated with clinical factors but such correlations may exist in larger samples. While it remains to be established whether FA alterations are a cause or a consequence of CPE—or both—they constitute an important step towards decrypting CPE.