Clinical Neurophysiology 112 (2001) 1761±1772
www.elsevier.com/locate/clinph
Cortical current density reconstruction of interictal epileptiform activity in temporal lobe epilepsy H.-J. Huppertz a,b,*, S. Hoegg a,b, C. Sick b, C.H. LuÈcking b, J. Zentner c, A. Schulze-Bonhage a, R. Kristeva-Feige b a Epilepsy Center, University of Freiburg, Breisacher Strasse 64, D-79106 Freiburg, Germany Department of Neurology, University of Freiburg, Breisacher Strasse 64, D-79106 Freiburg, Germany c Department of Neurosurgery, University of Freiburg, Breisacher Strasse 64, D-79106 Freiburg, Germany b
Accepted 4 May 2001
Abstract Objective: To investigate the value of cortical current density (CCD) reconstruction in localizing intracranial generators of interictal epileptiform activity in mesial and lateral temporal lobe epilepsy (TLE). Methods: Non-linear minimum L1-norm CCD reconstruction (with current sources restricted to the individual cortical surface and a realistic boundary element method (BEM) head model) was used to localize and to study the propagation of interictal epileptiform EEG activity in 13 pre-surgical patients with TLE. Results: In all but one patient with mesial temporal lesions, an initial activation maximum corresponding to the ascending part of averaged sharp waves was found in the ipsilateral anterior basolateral temporal lobe, mostly extending up to the affected mesial structures whose resection rendered the patients seizure-free. In all 3 patients with lateral temporal lesions, the activation was initially con®ned to temporal neocortex immediately adjacent to the epileptogenic lesion. Towards the peak of sharp waves, two patients showed a propagation of interictal activity to anterior and posterior and partly contralateral temporal regions. A conventional EEG analysis based on amplitude maxima or phase reversal would have missed the initial onset zone. Conclusions: The ®ndings demonstrate that CCD reconstruction can be a valuable additional non-invasive component in the multimodal pre-surgical evaluation of epilepsy patients. q 2001 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Epilepsy; Interictal epileptic activity; Electroencephalography; Magnetic resonance imaging; Source modeling; Current density reconstruction
1. Introduction In the past years, source modeling based on scalprecorded EEG has been increasingly used to localize epileptogenic foci in the pre-surgical evaluation of epilepsy patients. In the commonly applied equivalent current dipole model, synchronously activated cortical areas are represented by one or more dipolar electric model sources (Henderson et al., 1975; Cohen and Cuf®n, 1983; Scherg, 1990). From the distribution of scalp-recorded EEG potentials and the shapes and conductivities of the different head compartments (e.g. brain, skull, skin), an inverse solution is calculated for the location, orientation, and strength of possibly underlying intracranial sources. Recent advances include the co-registration of EEG data with anatomical * Corresponding author. Tel.: 149-761-270-5001; fax: 149-761-2705003. E-mail address:
[email protected] (H.-J. Huppertz).
information from magnetic resonance images (MRI) and the use of realistic head models (Gevins et al., 1990; Roth et al., 1993; Buchner et al., 1995; Kristeva-Feige et al., 1997; Grimm et al., 1998; Koles, 1998; Ball et al., 1999). Equivalent current dipole modeling based on EEG or magnetoencephalography (MEG) has been successfully used in localizing intracranial generators of interictal and ictal epileptiform activity (Ebersole, 1994; Baumgartner et al., 1995; Assaf and Ebersole, 1997; Boon et al., 1997; Krings et al., 1998; Diekmann et al., 1998). The results were in agreement with those of other imaging modalities (MRI, positron emission tomography (PET), single photon emission computed tomography (SPECT)) and intracranial EEG recordings (Nakasato et al., 1994; Stefan et al., 1994; Merlet et al., 1996; Roth et al., 1997; Shindo et al., 1998; Merlet and Gotman, 1999; Krings et al., 1999; Boon et al., 1999; Huppertz et al., 2001; Kobayashi et al., 2001). However, equivalent current dipole algorithms are a simpli®cation since they rely on the assumption that
1388-2457/01/$ - see front matter q 2001 Elsevier Science Ireland Ltd. All rights reserved. PII: S 1388-245 7(01)00588-0
CLINPH 2001511
1762
H.-J. Huppertz et al. / Clinical Neurophysiology 112 (2001) 1761±1772
synchronously activated cortical areas are well represented by their centers of mass and their mean surface normals. In addition, they require a priori knowledge of the number of active sources (Fuchs et al., 1999). Therefore, a growing emphasis has been on developing distributed source models as in current density reconstructions, which represent more realistically active brain regions, and which need no assumptions about the number, shape, or size of activated areas (Ebersole, 1999; Fuchs et al., 1999). Several methods have been proposed (Wang et al., 1992; Ilmoniemi, 1993; Pascual-Marqui et al., 1994; Hamalainen and Ilmoniemi, 1994; Phillips et al., 1997; Lantz et al., 1997; Uutela et al., 1999; Fuchs et al., 1999; Grave de Peralta Menendez et al., 2000). They all have in common a pre-determined distribution of elementary current dipoles on given positions in the head, either on a regular grid or constrained to the cortical surface. While the positions of these dipoles are ®xed, their orientations and strengths must be determined. The fact that the number of sensors or electrodes is usually much smaller than the number of modeling dipoles leads to a highly underdetermined problem, which can only be solved by additional constraints (e.g. minimum norm). The constraining model term has to be weighted against the data term by a regularization parameter (Ebersole, 1999; Fuchs et al., 1999). The present study was aimed at localizing cortical generators of interictal epileptiform activity in patients with temporal lobe epilepsy (TLE) by means of cortical current density (CCD) reconstruction as described recently by Fuchs et al. (1999). So far, there have been only a few reports about the application of this method for investigating somatosensory evoked potentials or localizing the source of epileptiform activity in up to 5 patients with TLE (Waberski et al., 1998, 1999, 2000). In this study, we wanted to evaluate if the method helps (1) to estimate the extension of active brain regions, (2) to investigate the timing and propagation of interictal epileptiform activity, and (3) to discriminate between mesial and lateral TLE. The results of source reconstruction were validated by comparison with the sites of the structural lesions, the post-operative outcome, and in two patients with the results of intracranial EEG recordings. In addition, the reconstruction results were compared to those of equivalent current dipole (ECD) modeling (Henderson et al., 1975; Cohen and Cuf®n, 1983; Scherg, 1990).
2. Methods 2.1. Patients Thirteen patients (7 males, 6 females, 13±64 years) with medically refractory epilepsy were selected from our video-EEG monitoring unit according to the following criteria: (1) presence of interictal epileptiform activity (spikes, sharp waves) in the EEG and (2) TLE according
to pre-surgical evaluation. All patients had a similar evaluation that included high-resolution MRI, interictal and ictal video-EEG, neuropsychological testing, in 8 cases interictal PET, in 7 cases interictal, in one case ictal SPECT, and in 3 cases magnetic resonance spectroscopy (MRS) of the hippocampi. Additional intracranial EEG recordings were necessary in two patients: patient 6 had non-lesional TLE and intrahippocampal depth electrode recordings established a seizure onset zone in the right anterior hippocampus. In patient 7, a temporomesial seizure onset zone could only be veri®ed by left-sided intrahippocampal depth electrodes and temporo-lateral and -basal strip electrodes. Post-operatively, 11 patients had excellent outcome, consistent with class I of the Engel classi®cation (mean follow-up, 250 days) (Engel et al., 1993). One patient showed a running-down phenomenon of the post-operative seizure frequency and has been seizure-free for 3 months now. Another patient (patient 7) still suffers from seizures whose semiology has changed, however, possibly due to an intracerebral hemorrhage during operation. However, he belonged to the two patients whose epileptogenic zone had been veri®ed pre-operatively by intracranial EEG recordings. In all patients, neuropathological analysis of the resected tissue was performed. The results of the pre-surgical evaluation and the post-operative outcome are summarized in Table 1. 2.2. EEG recordings All measurements were performed during pre-surgical video-EEG recording in our monitoring unit (Neuro®le NT EEG system). The EEGs were recorded from 23±31 scalp electrodes, placed according to the 10-20 system, with special coverage of the temporal lobes by a minimum of 4 additional sub-equatorial electrodes at anterior temporal (T1, T2) and mastoidal (MN1, MN2) positions (reference electrode on the forehead; electrode resistances below 10 kV, sampling rate 256 Hz, 16-bit resolution, bandpass ®lter 1.59±97 Hz). Additional bilateral sphenoidal electrodes (SP1, SP2) were used for identi®cation of epileptiform activity, but not for source reconstruction. Eye movements and blinks were monitored to exclude artifacts. Together with the EMG of the masseter muscle, they helped to de®ne sleep stages. Spikes and sharp waves were identi®ed visually according to IFSECN criteria (Chatrian et al., 1974) and marked provisionally on the highest negative peak in the EEG (after re-referentiating to a common average montage). Using a separate EEG analyzing software (Vision Analyzer, Brain Products), the marker positions were automatically adjusted to the earliest negative peak around the manually set markers. Artifact-free EEG epochs of 1000 ms before to 1000 ms after the marker positions were baseline-corrected and averaged. Epileptic discharges that differed by their
H.-J. Huppertz et al. / Clinical Neurophysiology 112 (2001) 1761±1772
1763
Table 1 Diagnoses, results of pre-surgical evaluation, and post-operative outcome of all patients a Patient Diagnosis (MRI/histology) Mesial temporal lesions 1 Right hippocampal sclerosis 2 Right hippocampal sclerosis 3 4 5
Left hippocampal sclerosis Left hippocampal sclerosis Left hippocampal sclerosis
6
No lesion in MRI, histologically signs of right hippocampal sclerosis Left hippocampal sclerosis
7
PET
SPECT
MRS
Operation
Outcome b
Right temporal pole Right temporal neocortex
No pathology Left TL
Both hippocampi (R . L)
AHE AHE
I I
No pathology
AHE AHE AHE
I I II c
Right mesial TL
AHE
I
Left mesial TL
AHE
IV d
Right TL 1 PL
ATLE
I
ATLE
I
Left mesial TL Left hippocampus 1 left superior temporal gyrus Right lateral TL and OL 1 left PL and OL Left TL and PL
Mesial temporal lesions extending to temporal neocortex 8 Right hippocampal sclerosis 1 focal cortical dysplasia of the temporal pole 9 Right hippocampal No pathology sclerosis 1 focal cortical dysplasia of the temporal pole 10 Glioneuronal hamartia in the right mesial TL including basal neocortex Lateral temporal lesions 11 Glioneuronal hamartia in the left inferior temporal gyrus 12 Ganglioglioma in the left basolateral TL 13 Dysembryoplastic neuroepithelial tumor in the right lateral TL
Left TL
No pathology
Both hippocampi (R . L)
LE (including AHE) I
Left inferior Right FL 1 TL TL 1 right and left FL
LE
I
LE
I
LE
I
a
FL, frontal lobe; TL, temporal lobe; PL, parietal lobe; OL, occipital lobe. AHE, selective amygdalohippocampectomy; ATLE, anterior temporal 2/3lobectomy; LE, lesionectomy. b According to Engel classi®cation. c Running-down phenomenon, seizure-free for the last 3 months. d New post-operative seizure type after intra-operative hemorrhage, seizures of old semiology disappeared.
waveform or scalp topography were averaged in separate groups. Baseline-correction and calculation of the signal-tonoise ratio (SNR) were based on the EEG data from 1000 to 200 ms before marker positions. Source reconstruction was applied to time points from 20 ms before to 12 ms after the peak of the averaged epileptiform discharges. For subsequent EEG/MRI co-registration, the electrode positions were digitized using an ultrasound localizing device (ZEBRIS) and the head contour was obtained by collecting approximately 1600±2000 points while moving the digitizer across the head surface. The transformation of electrode positions and MRI data into the same co-ordinate system was based on matching the digitized head surface with the head contour as segmented from MRI data by using an automatic surface matching technique (Huppertz et al., 1998). This procedure seems to be more reliable than matching a small number of reference points because the digitization error of single points largely averages out when
surfaces consisting of thousands of points are ®tted (Pelizzari et al., 1989; Kober et al., 1993). 2.3. Cortical current density reconstruction Source reconstruction was performed using non-linear CCD analysis. In contrast to linear CCD implementations minimizing the L2-norm of the reconstructed currents (minimum norm least squares, MNLS) (Ilmoniemi, 1991), the method calculates a regularized solution with a minimal sum of absolute current densities (minimum L1-norm). This non-linear CCD solution is more focal and of higher spatial resolution than the corresponding MNLS solution (Ball et al., 1999; Sick et al., 2000). In addition, it is more robust with regard to outliers in the measured data. Both effects result from the fact that large currents or data values are not punished by their squared strength as in the MNLS method, but contribute only with their (weighted) absolute
1764
H.-J. Huppertz et al. / Clinical Neurophysiology 112 (2001) 1761±1772
values to the function to be minimized (Wagner, 1998; Fuchs et al., 1999). To avoid over®tting, a reasonable CCD solution should not explain more of the data than that above the noise level. Thus, the regularization parameter l that weighted the constraining model term against the data term was determined by iterative adjustment until a residual deviation of the CCD solution of 1/SNR (with 5% tolerance) was achieved (x 2-criterion) (Wagner, 1998; Fuchs et al., 1999). Realistic 3-compartment boundary element method (BEM) head models with about 4000 (range 3600±4300) nodes per model were set up from the MRI data (MPRAGE sequence with TR/TE/alpha 9.7 ms, 4 ms, 128 on a 1.5 T Magnetom Vision, Siemens) by segmentation and triangulation of the 3 main compartments: brain with CSF (inside of skull), skull and skin (Wagner et al., 1995; Huppertz et al., 2001). The conductivities of the different tissues were de®ned as 0.33 S/m for the skin, 0.0042 S/m for the skull, and 0.33 S/m for the brain (Geddes and Baker, 1967). A separate compartment representing the cerebrospinal ¯uid was not de®ned because of potential mathematical errors arising from closely spaced boundary surfaces in a BEM model, in this case the surfaces of brain and cerebrospinal ¯uid (Fuchs et al., 1998). To include physiologic knowledge and the individual brain morphology, the CCD reconstructions were performed on about 28 000 supporting points distributed over the cortical surface as segmented from the individual MRI. The elementary current dipoles on these supporting points were restricted to be normal with respect to the cortical surface. To account for the depth dependency of the CCD reconstruction algorithm and to compensate for the lower gains of deeper dipole moments, the current locations were weighted by the inverse of the square root of the gains in the lead-®eld matrix (with locationwise singular value decomposition of the lead-®eld before gain determination). Compared to the usage of full gains as suggested by Fuchs et al. (1998), this depth normalization seems to provide more accurate solutions and avoids to overemphasize deep sources (Sick et al., 2000). Image segmentation, volume conductor modeling, source reconstruction, and visualization were performed using the CURRY software (Neurosoft, Inc.). Minimum Lp-norm with the norms of both the data and the model term set to 1 (thus equaling minimum L1-norm) were selected as the reconstruction mode. The regularization parameter l was adjusted manually according to the criterion described above. Only currents with at least 50% of the current strength at the maximum current density were considered. The reconstruction results, i.e. the elementary current dipoles were encoded as arrows of different sizes, colors, and orientations overlaying the segmented cortex. 2.4. Equivalent current dipole modeling In addition to CCD analysis, the same EEG data sets were
subjected to single ECD modeling (Henderson et al., 1975; Cuf®n, 1985; Scherg, 1990). The calculation was done with the CURRY software (Neurosoft, Inc.) and based on the same BEM head models as described above. The source space was de®ned as the compartment `brain with CSF' (inside of skull), reduced by 3 mm in a closing operation to avoid potential mathematical errors in the vicinity of boundary surfaces in a BEM model (Fuchs et al., 1998). The goodness of ®t (GoF) was determined as the percentage of EEG data explained by the solution of the ECD location. Only ECD results with a GoF . 95% were accepted and visualized in the MRI data sets. 3. Results In all patients, interictal epileptiform discharges ipsilateral to the side of their epileptogenic lesion could be recorded. In 4 patients (1, 9, 6, 4), an additional, relatively small sub-group (,20%) of sharp waves with their scalp maximum in temporal electrodes contralateral to the lesion was observed, mostly in sleep stages 1±2. Since it is known that during non-REM sleep, spike, and sharp waves tend to spread from the primary focus to ipsi- and contralateral brain regions (Sammaritano et al., 1991; Gigli and Valente, 2000), these data were not subjected to CCD analysis. Contralateral sharp waves were taken into account only in patient 11 who showed an almost even distribution of rightand left-sided temporal sharp waves. The ®nal averaged data included 29±212 (mean, 114) single epileptiform discharges per patient and sharp wave group. The SNR ranged from 3.4 to 29.7 (mean, 15.4) in the ascending part of the averaged sharp wave (at 220 ms) and from 7.6 to 67.4 (mean, 26.4) at the peak. This allowed source reconstructions with residual deviations of 3.7± 29.1% (mean, 12.5%) at 220 ms and of 1.5±13.1% (mean, 5.6%) at the sharp waves' peak. The results of CCD reconstruction are shown in Table 2. The patients were divided into 3 groups according to the site of their epileptogenic lesion. 3.1. Patients with mesial temporal lesions For both the ascending part and the peak of the averaged sharp waves, the CCD reconstruction showed in all 7 patients an activation of the basolateral temporal lobe ipsilateral to the epileptogenic lesion. In most cases, the activation was con®ned to anterior temporal regions and extended on the basal surface of the temporal lobe up to mesial structures (Fig. 1). Only in patients 1 and 7, sub-groups of sharp waves with amplitude maxima at T4 and MN1, respectively, were localized in the middle and posterior basolateral temporal lobe. In some patients, additional activations were found in the ipsilateral insula (e.g. patient 1), in the basal to polar frontal lobe (e.g. patients 2, 4, and 7), and in the contralateral temporal lobe (e.g. patients 4 and 7). However, these were
H.-J. Huppertz et al. / Clinical Neurophysiology 112 (2001) 1761±1772
1765
Table 2 Results of CCD reconstruction a Patient
Localization of earliest sharp wave CCD localization (and residual deviation) at: peak 1 number of sharp waves (SNR at peak) 220 ms 212 ms
Mesial temporal lesions 1 T2 97 SWs (34.1)
T4 86 SWs (26.0)
Basolateral anterior TL 1 slight activation of insula and basal FL (4.0%) Basolateral middle TL 1 slight activation of insula (5.2%) Basal to polar FL 1 basolateral anterior TL (14.1%) Basolateral anterior TL (11.4%)
2
Fp2 95 SWs (17.5)
3
SP1 184 SWs (12.2)
4
F7 29 SWs (8.3)
5
SP1 184 SWs (22.9)
6
F8 36 SWs (7.6)
7
F7 106 SWs (37.2)
(29.1%) Basal and lateral anterior TL 1 adjacent basal FL (3.7%)
MN1 53 SWs (34.2)
Basolateral posterior TL
0 ms (peak)
112 ms
!
!
!
(3.6%) !
(2.8%) !
(3.4%) !
(4.3%) !
(3.8%) Basolateral anterior TL 1 slight activation of basal to polar FL (6.0%) ! (8.5%) !
(3.4%) !
(12.5%) ! (4.6%) Basolateral anterior TL 1 slight activation of posterolateral TL (13.1%) !
(12.5%) ! (4.0%) !
(8.6%) ! (9.3%) Minimal activation of basolateral Basolateral anterior TL anterior TL and Fl 1 contralateral TL (25.7%) (16.0%) Basolateral anterior TL ! (10.9%) (6.5%) Basolateral anterior TL !
(5.5%)
(17.5%) !
(2.7%) (2.6%) Basolateral posterior Basolateral posterior TL 1 slight activation of FL TL 1 slight activation of FL and contralateral TL (4.1%) (3.1%)
Mesial temporal lesions extending to temporal neocortex 8 T4 126 SWs (35.1) Basolateral anterior TL ! (4.6%) (3.4%) 9 F8 212 SWs (67.4) Basolateral middle to anterior TL ! (3.9%) (2.1%) 10 SP2 139 SWs (9.6) Basolateral anterior ! TL 1 FL 1 contralateral temporal pole (12.9%) (12.2%) Lateral temporal lesions 11 SP1 103 SWs (23.5)
SP2 137 SWs (28.7)
12
F7 57 SWs (17.5)
13
SP2 182 SWs (40.4)
Middle inferior TL ± immediately adjacent to the lesion (10.1%) Contralateral anterolateral TL
Middle inferior TL ± adjacent to the lesion (8.3%) Contralateral anterior basolateral TL
(15.7%) ! (3.3%) Basal and lateral anterior TL 1 basal FL 1 contralat. TL (2.9%)
! (2.7%) ! (1.5%) !
! (3.2%) ! (1.6%) !
(10.7%)
(11.9%)
Anterolateral and posterior TL
Middle inferior TL ± adjacent to the lesion (4.1%) !
(11.8%) !
(5.6%) Contralateral anterior basolateral TL 1 slight activation of basal FL (3.5%) (3.9%) Basolateral anterior and middle Basolateral anterior to inferior TL 1 contralateral TL posterior TL 1 contralateral TL (6.0%) (6.9%) ! !
(3.4%)
(2.6%)
(27.7%) (8.1%) Middle inferior TL ± immediately ! under the lesion (28.0%) Lateral and slightly basal activation of TL ± immediately anterior to the lesion (3.9%)
(5.5%) ! (8.7%) !
(2.7%)
a
FL, frontal lobe; TL, temporal lobe; PL, parietal lobe; OL, occipital lobe; SWs, sharp waves; ! , unchanged CCD localization compared to the preceding time point.
only weak activations, the activation maxima were still in the basolateral temporal lobe. Only in patient 2, the analysis of sharp waves at 220 ms yielded an activation in the basal to polar frontal lobe that exceeded the temporal activation. At the peak of the averaged sharp waves, the CCD recon-
struction showed again a maximum in the anterior basolateral temporal lobe. Over the time course of 220 to 112 ms around the peak of averaged sharp waves, the source reconstructions displayed only slight propagation of interictal epileptiform
1766
H.-J. Huppertz et al. / Clinical Neurophysiology 112 (2001) 1761±1772
In patient 10, the activation was not con®ned to this region, but included to a lesser degree the ipsilateral frontal lobe and the contralateral temporal pole. Towards the sharp wave peak, the activation focussed on the anterior basolateral temporal lobe with mesial extension up to the border of the glioneuronal hamartia. 3.3. Patients with lateral temporal lesions In all 3 patients, the CCD reconstruction for both the ascending part and the peak of averaged sharp waves showed an activation close to the epileptogenic lesion. For patient 13, the activation remained exclusively con®ned to the close vicinity of the lesion over the investigated time course. In contrast, patients 11 and 12 showed a propagation and extension of interictal epileptiform activity from an inferior mid-temporal location (immediately adjacent to their lesion) to anterior and posterior temporal regions, and in patient 12 also to the contralateral temporal lobe (Figs. 2 and 3). As mentioned above, about half of the sharp waves in patient 11 had their initial amplitude maximum in the contralateral sphenoidal electrode. Their source reconstruction showed an activation of the contralateral antero- and basolateral temporal lobes. 3.4. Comparison between the results of CCD reconstruction and ECD modeling Fig. 1. CCD reconstruction results for the peak of averaged sharp waves in 4 patients with hippocampal sclerosis. The elementary current dipoles are encoded as arrows of different sizes, colors, and orientations overlaying the segmented cortex, which is shown in side, frontal, and bottom view. The affected hippocampi were segmented separately and are shown in brown color. For each patient, the residual deviation and the current maximum (i.e. the maximum dipole moment per area) of the CCD solution are displayed. The current distributions were clipped at 50% of the maximum current density.
activity in this patient group. In patient 7, for example, the analysis of a sub-group of sharp waves with amplitude maxima at MN1 showed a shift of the activation from posterior to anterior basolateral temporal regions and an extension to the ipsilateral frontal and contralateral temporal lobe. In patient 6, the activation extended from anterior to posterior basolateral temporal regions. As mentioned, the activation maximum in patient 2 shifted from the ipsilateral frontal to the anterior basolateral temporal lobe. In the other cases, the CCD reconstruction results for the ascending part and the peak of sharp waves differed only in terms of current strengths, which increased towards the peak. 3.2. Patients with mesial temporal lesions extending to temporal neocortex For all patients in this group, the CCD reconstructions over the investigated time course of 220 to 112 ms yielded activation maxima in the anterior or middle to anterior basolateral temporal lobe ipsilateral to the epileptogenic lesion.
In contrast to the extended results of CCD reconstruction, the point-like dipole locations of single ECD modeling could be determined with gyral or even subgyral accuracy. They are shown in Table 3 together with the corresponding GoF values. In most cases, the ECD results coincided well with the activation maxima of the CCD solutions. In patients with mesial temporal lesions, e.g. dipole locations in the anterior part of the inferior temporal or the parahippocampal gyrus (e.g. patients 2, 3, 5, and 7) corresponded to initial CCD activation maxima in the anterior basolateral temporal lobe. When the CCD solutions showed activations of the posterior temporal lobe (e.g. sharp waves at T4 in patient 1 or at MN1 in patient 7), the ECD results were accordingly shifted to middle or posterior parts of the basolateral temporal lobe. In patient 1 where the CCD reconstruction indicated additional involvement of insula and basal frontal lobe for sharp waves at T2, the dipole was located in the middle of the different activation areas, i.e. in the anterior to middle part of the superior temporal gyrus. A similar situation was found in patient 10 where CCD reconstruction showed extended activation in the basolateral anterior temporal, basal frontal, and contralateral temporal lobe while the dipole location was found in the posterior part of ipsilateral orbital gyri. It should be noticed that the ECD results in patient 1 were accompanied by high GoF values ranging from 97 to almost 99%. Generally, this indicates that applying a single dipole model was adequate and only a limited cortex area was activated.
H.-J. Huppertz et al. / Clinical Neurophysiology 112 (2001) 1761±1772
Fig. 2. CCD reconstruction results of a patient with a glioneuronal hamartia in the left inferior temporal lobe (patient 11, lesion shown in brown color). Sharp waves with amplitude maxima at SP1 were averaged and investigated from 20 ms before to 12 ms after the peak to study the propagation of interictal epileptiform activity. For each time point, the residual deviation and the current maximum of the CCD solution are displayed.
Within the group of patients with lateral temporal lesions, ECD and CCD reconstruction results totally coincided for patient 13 and for the sharp waves at SP2 in patient 11. However, for the sharp waves at SP1 and F7 in patients 11 and 12, respectively, the ECD results were located more anteriorly in the inferior temporal lobe as compared to the CCD solution. Especially for the ascending part of sharp waves, the dipoles were located signi®cantly anterior to the epileptogenic lesion and failed to show the propagation of interictal activity from the vicinity of the lesions to other temporal areas. Finally, it is noteworthy that for quite a number of patients, reliable ECD results are missing in Table 3 because of unacceptably low GoF values, especially for the ascending part of the averaged sharp waves (i.e. at 220 ms) and in case of generally low SNR (e.g. patients 4, 6, and 10). 4. Discussion The present study investigated the value of CCD reconstruction in localizing generators of interictal epileptiform activity in patients with TLE. No prior assumptions about
1767
Fig. 3. Propagation of interictal epileptiform activity in a patient with a ganglioglioma in the left basolateral temporal lobe and sharp waves with amplitude maxima at F7 (patient 12). Over the time course of the averaged sharp waves, the interictal epileptiform activity propagates from an inferior mid-temporal location immediately adjacent to the lesion to anterior and posterior temporal regions and also to the contralateral temporal lobe.
the number and location of sources were made, except that all sources were constrained to a surface representing the individual cortical gray matter. The results were validated by comparison with the sites of the structural lesions, whose epileptogenic nature was con®rmed by an excellent postoperative outcome in 12 patients, and the results of intracranial EEG recordings in the remaining patient. In all but one patients with mesial temporal lesions, an initial activation maximum corresponding to the ascending part of averaged sharp waves was found in the anterior basolateral temporal lobe ipsilateral to the side of the epileptogenic lesion. On the basal surface of the temporal lobe, this activation mostly extended up to mesial structures, i.e. the hippocampus and amygdala. Only in patient 2, the localization results indicated that initially frontal areas were predominantly involved in the generation of interictal epileptiform activity before the activation maximum shifted and focussed in the anterior basolateral temporal lobe for the peak of the averaged sharp waves. One could speculate that via fasciculus uncinatus, the affected hippocampus ®rst in¯uenced fronto-basal areas before adjacent temporal neocortical tissue was involved.
1768
H.-J. Huppertz et al. / Clinical Neurophysiology 112 (2001) 1761±1772
Table 3 Results of ECD localization a Patient
Localization of earliest sharp wave peak 1 number of sharp waves (SNR at peak)
ECD localization (and GoF) at:
220 ms
212 ms
0 ms (peak)
112 ms
Anterior to middle part of superior temporal gyrus (98.01%) Middle part of inferior temporal gyrus (98.2%) Anterior part of inferior temporal gyrus (97.0%) Inferior temporal gyrus, near temporal pole (97.9%) ± (94.1%) Anterior part of inferior temporal gyrus
!
!
!
(97.3%) Middle part of superior temporal gyrus (98.2%) Temporal pole
(97.0%) Transverse temporal gyri ± insula (98.0%) !
(98.9%) !
(97.4%) !
(97.9%) !
(98.1%) Temporal pole (95.0%) !
(98.4%) ± (90.7%) !
(97.9%) Temporal pole (95.8%) Medial occipitotemporal gyrus, close to the pes of hippocampus (98.5%) ± (92.7%) Temporal pole
(98.0%) !
(98.2%) !
(96.5%)
(97.1%)
(98.3%) Anterior part of parahippocampal gyrus (97.6%)
!
!
!
(98.7%) !
(98.7%) !
(98.9%) ± (91.8%)
(99.2%) Posterior part of orbital gyri (95.3%)
(99.3%) ± (94.8%)
(98.5%) Anterior part of parahippocampal gyrus (99.0%) Posterior part of orbital gyri (95.2%)
SP2 137 SWs (28.7)
Anterior part of inferior temporal gyrus (97.2%) ±
Transverse temporal gyri ± insula (95.9%) Anterior part of contralateral superior temporal gyrus
12
F7 57 SWs (17.5)
(81.2%) ±
Middle part of superior temporal gyrus (96.1%) Contralateral inferior temporal gyrus, near temporal pole (97.0%) Anterior part of inferior temporal gyrus
13
SP2 182 SWs (40.4)
Mesial temporal lesions 1 T2 97 SWs (34.1)
T4 86 SWs (26.0)
2
Fp2 95 SWs (17.5)
3
SP1 184 SWs (12.2)
4
F7 29 SWs (8.3)
5
SP1 184 SWs (22.9)
6
F8 36 SWs (7.6)
7
F7 106 SWs (37.2)
MN1 53 SWs (34.2)
(97.8%) ± (86.0%) Anterior part of parahippocampal gyrus (97.7%) Posterior part of parahippocampal gyrus (96.1%)
Mesial temporal lesions extending to temporal neocortex 8 T4 126 SWs (35.1) Anterior part of medial occipito-temporal gyrus (98.1%) 9 F8 212 SWs (67.4) Temporal pole
10
SP2 139 SWs (9.6)
Lateral temporal lesions 11 SP1 103 SWs (23.5)
(67.0%) Temporal pole ± immediately anterior to the lesion (99.0%)
(97.5%) Anterior part of superior temporal gyrus (97.9%) ! (98.2%) ± (94.5%) !
(98.5%) ± (91.8%) !
! (95.4%) !
(97.1%) !
(97.7%) !
(97.4%) Anterior part of inferior temporal to lateral occipitotemporal gyrus (98.3%) !
(98.9%)
(98.8%)
(98.5%)
(98.3%) !
a
SWs, sharp waves. ±, dipole localization not accepted because of low GoF (GoF , 95%).; ! , unchanged dipole localization compared to the preceding time point.
No patient showed an activation con®ned to the mesial temporal lobe. In all cases, the CCD localizations primarily comprised basolateral temporal regions and only extended up to mesial structures. This is in line with the results of a previous study where equivalent current dipole localizations
did not appear in the epileptogenic hippocampi, but in adjacent temporal neocortex, i.e. the inferior temporal or medial occipitotemporal gyrus (Huppertz et al., 2001). The reason could be that in cases of mesial temporal lesions, epileptiform activity generated in the temporobasal or temporolat-
H.-J. Huppertz et al. / Clinical Neurophysiology 112 (2001) 1761±1772
eral cortex can mask activity generated in deeper structures, or that epileptiform activity is not `visible' for surface electrodes until it has propagated to adjacent neocortex. The lack of identi®able scalp potentials when activation is con®ned to the hippocampus has been explained by the relatively small activated tissue volume and the curved geometry causing external ®eld cancellation (Lopes da Silva and Van Rotterdam, 1993; Pacia and Ebersole, 1997). According to Alarcon et al. (1994), physiologically unrealistic voltage gradients would be necessary for deep sources to evoke detectable scalp potentials. This view is supported by ®ndings of Merlet and Gotman (1999) who compared dipole locations based on scalp EEG paroxysms with intracerebral potentials recorded simultaneously. They never observed scalp EEG spikes corresponding to focal activity limited to mesial temporal structures. The lateral temporal neocortex was always involved. Recently, the same authors compared source localizations of ictal epileptic activity with intracerebral EEG and also concluded that mesial temporal seizure discharges did not contribute to scalp EEG activity (Merlet and Gotman, 2001). Therefore, the ability to detect and localize pure mesial epileptic activity perhaps does not depend on the choice of the source reconstruction technique, whether it is CCD reconstruction, low-resolution electromagnetic tomography (LORETA) (Pascual-Marqui et al., 1994; Lantz et al., 1997) or synthetic aperture magnetometry (SAM) (Robinson and Vrba, 1999), but is generally limited due to anatomical and physiological reasons. In all 3 patients with lateral temporal lesions, the present study showed an initial activation of a small temporal region immediately adjacent to the lesion whose later resection rendered the patient seizure-free. In one patient, the activation remained con®ned to the close vicinity of the lesion while the others displayed a propagation of interictal activity to anterior and posterior and partly contralateral temporal regions. An analysis restricted to the peak of their sharp waves would have failed to localize the onset of interictal activity close to their lesions (Figs. 2 and 3). Although this study was not planned as a systematic comparison of electric source reconstruction to other imaging techniques like PET, SPECT, or MRS, it should be mentioned that on a lobar level, the CCD reconstruction of interictal epileptiform activity in our patients was more in accordance with the location of the epileptogenic lesions than the results of the aforementioned methods. Three of 8 SPECT and one of 3 MRS investigations showed no pathology or gave falsely lateralizing results, and 4 of 8 PET ®ndings were of little localizing value, i.e. displayed interictal hypometabolism in more than one lobe (Table 1). The ®ndings of this study demonstrate that the novel approach of CCD reconstruction is able to localize correctly cortical areas involved in the generation of interictal epileptiform activity. In contrast to source reconstructions based on dipolar source models, which characterize only the center of mass of activated brain regions, the CCD recon-
1769
struction gives an impression how far the activation extends and helps to delineate the irritative zone. Especially in case of extended irritative areas involving more than one lobe, a single dipole solution can give a misleading impression of focality and the dipole can be shifted to unexpected locations, e.g. the superior temporal gyrus for the sharp waves at T2 in patient 1 or the orbital gyri in patient 10 who both had mesial temporal lesions. Fitting a second or third dipole might have helped to visualize additionally active brain areas, but relatively high GoF values of 97±99% at least in patient 1 did not give ground for such an approach. However, to what extent the results of CCD reconstruction exactly overlap with the irritative zone cannot be derived from this study, but has to be determined by simulations or studies in patients with implanted grid electrodes. As shown by the results of patients 11 and 12, the method of CCD reconstruction is also able to investigate the propagation of interictal activity. A conventional EEG analysis based on amplitude maxima or phase reversal would have attributed their sharp waves at F7 and SP1, respectively, to the anterior temporal lobe and would have missed the midtemporal onset zone adjacent to the lesion. These cases underline the importance of analyzing the whole time course and not only the peak of sharp waves, which may be dominated by propagated activity (Scherg et al., 1999). Interestingly, the results of dipole modeling for the ascending part of averaged sharp waves in these two patients were also located in the anterior inferior temporal lobe, i.e. near the electrode contact with the highest amplitude and did not visualize the onset zone near the mid-temporal lesion. This indicates that in some cases, CCD reconstruction can be superior to single dipole modeling, possibly due to the inclusion of source orientation as an additional constraint. The discrimination of mesial and lateral TLE by CCD reconstruction is limited by the aforementioned fact that epileptiform activity arising from mesial structures does not seem to become visible for surface electrodes until propagation to adjacent neocortical tissue. However, this handicap is valid for all source reconstruction methods. In addition, the results of this study, although derived from a limited number of patients so far, indicate that the involvement of basal temporal regions during the ascending part of sharp waves can be a hint to a mesial origin of the epileptiform activity. In patient 7, for example, the CCD reconstruction could have helped to avoid invasive recording if one had interpreted the early involvement of basal temporal regions in both sharp wave groups (F7 and MN1) as a reliable marker of a mesially located epileptogenic zone. Several methodological problems associated with CCD reconstruction should be mentioned. First of all, a suf®cient SNR is mandatory for reliable reconstruction results. Although the non-linear L1-norm has been reported to perform better than linear methods such as MNLS or LORETA (Pascual-Marqui et al., 1994; Lantz et al., 1997) with respect to localization accuracy and spatial resolution, the L1-norm algorithm suffers most from low SNRs by
1770
H.-J. Huppertz et al. / Clinical Neurophysiology 112 (2001) 1761±1772
showing scattered and spurious results for deep source locations (Fuchs et al., 1999). But what is a reasonable SNR below which reconstruction results are no longer trustworthy? In the present study, SNR values ranged from 3.4 to 67 (average, 15.4 for 220 ms and 26.4 for the peak of sharp waves). For SNR values below 8 (mostly at 220 ms in the ascending part of sharp waves), both very focussed (e.g. patients 6 and 12 with SNR 3.4 and 3.5, respectively) and somewhat distributed, but nevertheless reasonable reconstruction results (e.g. patients 4 and 10 with SNR 3.9 and 7.4, respectively) were found. Thus, there is probably no de®nite cut-off value but an SNR range of 8±3 where localization results become more and more questionable and should be validated by other imaging methods. However, the present study is not adequate to answer this question in detail. Again, simulation studies with different SNR values and sources in different depths are necessary to solve this problem (Sick et al., in preparation). The problem of choosing a correct regularization parameter is closely related to the SNR of the data. In contrast to source reconstruction based on one or two equivalent current dipoles, small residual deviations between measured and calculated EEG potentials are no longer a good criterion for the quality of the CCD reconstruction results. They can always be achieved by selecting small values for the regularization parameter, i.e. down-weighting the model term. In case of over®tting the data by using a wrong regularization parameter, the algorithm starts to ®t the noise or correlated background features of the data, which leads to spurious results or ghost sources. Therefore, it is important to estimate the noise in the data and adjust the regularization parameter accordingly (Fuchs et al., 1999). The computational effort is high since adjusting the regularization parameter has to be done iteratively. With almost 30 000 cortical supporting points in our patients, each iteration required about 4±5 min on a PC with Pentium III 600 MHz CPU. Usually, 4±8 iterations were necessary until the residual deviation of the CCD solution matched the inverse of the SNR of the data. Since the automatic algorithm did not reliably converge, the new value for the regularization parameter had to be chosen manually for each iteration. However, these problems will hopefully be overcome by the development of better performing CPUs and reliable automatic algorithms for the adjustment of the regularization parameter. In the present study, the elementary current dipoles on the cortical supporting points were restricted to be normal with respect to the cortical surface. Since including the surface normals punishes cortical supporting points with wrong normal orientation, the results sometimes appeared somewhat fragmented for lower SNR values (e.g. current dipoles of signi®cant strength were found only on top of the gyri, but not in adjacent sulci or vice versa), while for the same data, the CCD reconstruction with rotating current dipoles showed coherent activation areas. Nevertheless, the more constrained source model with surface normals was chosen
because with rotating current dipoles, the activation was usually con®ned to the cortical area immediately below the surface electrode showing the highest amplitude for the investigated time point. Only the inclusion of surface normals forced the algorithm to adequately emphasize more distant, but correctly orientated areas. This may also explain the difference to the recent results of Waberski et al. (2000) who found no advantage of current density reconstruction compared to single moving dipole or spatio-temporal dipole modeling. They used the same reconstruction technique with a non-linear L1-norm formulation except that the calculation was done with a smaller number of cortical supporting points (15 000±18 000) and, above all, without a constraint for source orientation. In conclusion, the results of this study indicate that the method of CCD reconstruction is able to reliably identify brain regions involved in the generation of interictal epileptiform activity. In particular, it may help to localize the cortical area generating the earliest part of the epileptiform discharge, to study the propagation, and to assess the extension of the irritative zone. It can indicate the epileptogenic nature of a structural lesion or guide the placement of intracranial electrodes when necessary. On the other hand, it may increase con®dence in the localization of the epileptogenic zone and thereby obviate the need for invasive recordings. Compared to equivalent current dipole modeling, CCD reconstruction has the advantage that no prior assumptions about the number of active sources are required. The method seems to be a valuable additional non-invasive component in the multimodal pre-surgical evaluation of epilepsy patients. Acknowledgements The study was partly supported by grants from the Deutsche Forschungsgemeinschaft (KR 1392/7-1) and the Research Fund of the Albert-Ludwigs-University Freiburg. References Alarcon G, Guy CN, Binnie CD, Walker SR, Elwes RD, Polkey CE. Intracerebral propagation of interictal activity in partial epilepsy: implications for source localisation. J Neurol Neurosurg Psychiatry 1994;57:435±449. Assaf BA, Ebersole JS. Continuous source imaging of scalp ictal rhythms in temporal lobe epilepsy. Epilepsia 1997;38(10):1114±1123. Ball T, Schreiber A, Feige B, Wagner M, LuÈcking CH, Kristeva-Feige R. The role of higher-order motor areas in voluntary movement as revealed by high-resolution EEG and fMRI. Neuroimage 1999;10:682±694. Baumgartner C, Lindinger G, Ebner A, Aull S, Serles W, Olbrich A, Lurger S, Czech T, Burgess R, LuÈders HO. Propagation of interictal epileptic activity in temporal lobe epilepsy. Neurology 1995;45(1):118±122. Boon P, D'Have M, Adam C, Vonck K, Baulac M, Vandekerckhove T, De Reuck J. Dipole modeling in epilepsy surgery candidates. Epilepsia 1997;38(2):208±218. Boon P, D'Have M, Van Hoey G, Vanrumste B, Vonck K, Adam C, Vandekerckhove T, Michielsen G, Baulac M, De Reuck J. Source localization in refractory partial epilepsy. Rev Neurol (Paris) 1999;155:499±508.
H.-J. Huppertz et al. / Clinical Neurophysiology 112 (2001) 1761±1772 Buchner H, Waberski TD, Fuchs M, Wischmann HA, Wagner M, Drenckhahn R. Comparison of realistically shaped boundary-element and spherical head models in source localization of early somatosensory evoked potentials. Brain Topogr 1995;8(2):137±143. Chatrian GE, Bergamini L, Dondey M, Klass DW, Lennox-Buchtal M, Petersen I. A glossary of terms most commonly used by clinical electroencephalographers. Electroenceph clin Neurophysiol 1974;37:538± 548. Cohen D, Cuf®n BN. Demonstration of useful differences between magnetoencephalogram and electroencephalogram. Electroenceph clin Neurophysiol 1983;56:38±51. Cuf®n BN. A comparison of moving dipole inverse solutions using EEGs and MEGs. IEEE Trans Biomed Eng 1985;32:905±910. Diekmann V, Becker W, JuÈrgens R, Grozinger B, Kleiser B, Richter HP, Wollinsky KH. Localisation of epileptic foci with electric, magnetic and combined electromagnetic models. Electroenceph clin Neurophysiol 1998;106:297±313. Ebersole JS. Non-invasive localization of the epileptogenic focus by EEG dipole modeling. Acta Neurol Scand 1994;152:20±28. Ebersole JS. EEG source modeling. The last word. J Clin Neurophysiol 1999;16:297±302. Engel Jr. J, Van Ness PC, Rasmussen TB, Ojemann LM. Outcome with respect to epileptic seizures. In: Engel J, editor. Surgical treatment of the epilepsies, New York, NY: Raven Press, 1993. pp. 609±621. Fuchs M, Wischmann HA, Wagner M, Theissen A. Performance of realistically shaped boundary element method volume conductor models. BIOMAG98 Abstracts, 1998;86. Fuchs M, Wagner M, Kohler T, Wischmann HA. Linear and nonlinear current density reconstructions. J Clin Neurophysiol 1999;16:267±295. Geddes LA, Baker LE. The speci®c resistance of biological material ± a compendium of data for the biomedical engineer and physiologist. Med Biol Eng 1967;5:271±293. Gevins A, Brickett P, Costales B, Le J, Reutter B. Beyond topographic mapping: towards functional-anatomical imaging with 124-channel EEGs and 3-D MRIs. Brain Topogr 1990;3:53±64. Gigli GL, Valente M. Sleep and EEG interictal epileptiform abnormalities in partial epilepsy. Clin Neurophysiol 2000;111(Suppl 2):60±64. Grave de Peralta Menendez R, Gonzalez Andino SL, Morand S, Michel CM, Landis T. Imaging the electrical activity of the brain: ELECTRA. Hum Brain Mapp 2000;9(1):1±12. Grimm C, Schreiber A, Kristeva-Feige R, Mergner T, Hennig J, LuÈcking CH. A comparison between electric source localisation and fMRI during somatosensory stimulation. Electroenceph clin Neurophysiol 1998;106(1):22±29. Hamalainen MS, Ilmoniemi RJ. Interpreting magnetic ®elds of the brain: minimum norm estimates. Med Biol Eng Comput 1994;32(1):35±42. Henderson CJ, Butler SR, Glass A. The localization of equivalent dipoles of EEG sources by the application of electrical ®eld theory. Electroenceph clin Neurophysiol 1975;39:117±130. Huppertz HJ, Hof E, Klisch J, Wagner M, LuÈcking CH, Kristeva-Feige R. Localization of interictal delta and epileptiform EEG activity associated with focal epileptogenic brain lesions. Neuroimage 2001;13(1):15±28. Huppertz HJ, Otte M, Grimm C, Kristeva-Feige R, Mergner T, LuÈcking CH. Estimation of the accuracy of a surface matching technique for registration of EEG and MRI data. Electroenceph clin Neurophysiol 1998;106:409±415. Ilmoniemi RJ. Estimates of neuronal current distributions. Acta Otolaryngol Suppl Stockh 1991;491:80±86. Ilmoniemi RJ. Models of source currents in the brain. Brain Topogr 1993;5(4):331±336. Kobayashi K, Merlet I, Gotman J. Separation of spikes from background by independent component analysis with dipole modeling and comparison to intracranial recording. Clin Neurophysiol 2001;112(3):405±413. Kober H, Grummich P, Vieth J. Precise fusion of MEG and MRI tomography using a surface ®t. Biomed Engineer (Berlin) 1993;38(Suppl): 355±356.
1771
Koles ZJ. Trends in EEG source localization. Electroenceph clin Neurophysiol 1998;106:127±137. Krings T, Chiappa KH, Cuf®n BN, Buchbinder BR, Cosgrove GR. Accuracy of electroencephalographic dipole localization of epileptiform activities associated with focal brain lesions. Ann Neurol 1998;44(1): 76±86. Krings T, Chiappa KH, Cuf®n BN, Cochius JI, Connolly S, Cosgrove GR. Accuracy of EEG dipole source localization using implanted sources in the human brain. Clin Neurophysiol 1999;110:106±114. Kristeva-Feige R, Grimm C, Huppertz HJ, Otte M, Schreiber A, JaÈger D, Feige B, BuÈchert M, Hennig J, Mergner T, LuÈcking CH. Reproducibility and validity of electric source localisation with high-resolution electroencephalography. Electroenceph clin Neurophysiol 1997;103(6): 652±660. Lantz G, Michel CM, Pascual-Marqui RD, Spinelli L, Seeck M, Seri S, Landis T, Rosen I. Extracranial localization of intracranial interictal epileptiform activity using LORETA (low resolution electromagnetic tomography). Electroenceph clin Neurophysiol 1997;102:414±422. Lopes da Silva F, Van Rotterdam A. Biophysical aspects of EEG and magnetoencephalogram generation. In: Niedermeyer E, Lopes da Silva F, editors. Electroencephalography. Basic principles, clinical applications, and related ®elds, Baltimore, MD: Williams and Wilkins, 1993. pp. 78±91. Merlet I, Gotman J. Reliability of dipole models of epileptic spikes. Clin Neurophysiol 1999;110:1013±1028. Merlet I, Gotman J. Dipole modeling of scalp electroencephalogram epileptic discharges: correlation with intracerebral ®elds. Clin Neurophysiol 2001;112(3):414±430. Merlet I, Garcia-Larrea L, Gregoire MC, Lavenne F, Mauguiere F. Source propagation of interictal spikes in temporal lobe epilepsy. Correlations between spike dipole modelling and [18F]¯uorodeoxyglucose PET data. Brain Pt 2 1996;119:377±392. Nakasato N, Levesque MF, Barth DS, Baumgartner C, Rogers RL, Sutherling WW. Comparisons of MEG, EEG, and ECoG source localization in neocortical partial epilepsy in humans. Electroenceph clin Neurophysiol 1994;91(3):171±178. Pacia SV, Ebersole JS. Intracranial EEG substrates of scalp ictal patterns from temporal lobe foci. Epilepsia 1997;38(6):642±654. Pascual-Marqui RD, Michel CM, Lehmann D. Low resolution electromagnetic tomography: a new method for localizing electrical activity in the brain. Int J Psychophysiol 1994;18(1):49±65. Pelizzari CA, Chen GT, Spelbring DR, Weichselbaum RR, Chen CT. Accurate three-dimensional registration of CT, PET, and/or MR images of the brain. J Comput Assisted Tomogr 1989;13:20±26. Phillips JW, Leahy RM, Mosher JC. MEG-based imaging of focal neuronal current sources. IEEE Trans Med Imaging 1997;16:338±348. Robinson SE, Vrba J. Functional neuroimaging by synthetic aperture magnetometry. In: Yoshimoto T, Kotani M, Karibe H, Nakasato N, editors. Recent advances in biomagnetism, Sendai: Tohoku University Press, 1999. pp. 302±305. Roth BJ, Balish M, Gorbach A, Sato S. How well does a three-sphere model predict positions of dipoles in a realistically shaped head? Electroenceph clin Neurophysiol 1993;87(4):175±184. Roth BJ, Ko D, von Albertini-Carletti IR, Scaf®di D, Sato S. Dipole localization in patients with epilepsy using the realistically shaped head model. Electroenceph clin Neurophysiol 1997;102(3):159±166. Sammaritano M, Gigli GL, Gotman J. Interictal spiking during wakefulness and sleep and the localization of foci in temporal lobe epilepsy. Neurology 1991;41:290±297. Scherg M. Fundamentals of dipole source potential analysis. In: Hoke M, editor. Advances in audiology, Basel: Karger, 1990. pp. 40±69. Scherg M, Bast T, Berg P. Multiple source analysis of interictal spikes: goals, requirements, and clinical value. J Clin Neurophysiol 1999;16: 214±224. Shindo K, Ikeda A, Musha T, Terada K, Fukuyama H, Taki W, Kimura J, Shibasaki H. Clinical usefulness of the dipole tracing method for localizing interictal spikes in partial epilepsy. Epilepsia 1998;39:371±379.
1772
H.-J. Huppertz et al. / Clinical Neurophysiology 112 (2001) 1761±1772
Sick C, LuÈcking CH, Wiesendanger M, Kristeva-Feige R. Comparison of different cortical current density methods for electric source reconstruction. Proceedings of the 30th Annual Meeting of the Society for Neuroscience, New Orleans, LA, Nov 4±9, 2000. p. 2203. Stefan H, SchuÈler P, Abraham-Fuchs K, Schneider S, Gebhardt M, Neubauer U, Hummel C, Huk WJ, Thierauf P. Magnetic source localization and morphological changes in temporal lobe epilepsy: comparison of MEG/EEG, ECoG and volumetric MRI in presurgical evaluation of operated patients. Acta Neurol Scand 1994;152(Suppl):83±88. Uutela K, Hamalainen M, Somersalo E. Visualization of magnetoencephalographic data using minimum current estimates. Neuroimage 1999;10:173±180. Waberski TD, Buchner H, Lehnertz K, Hufnagel A, Fuchs M, Beckmann R, Rienacker A. Properties of advanced headmodelling and source reconstruction for the localization of epileptiform activity. Brain Topogr 1998;10(4):283±290. Waberski TD, Buchner H, Perkuhn M, Gobbele R, Wagner M, Kucker W,
Silny J. N30 and the effect of explorative ®nger movements: a model of the contribution of the motor cortex to early somatosensory potentials. Clin Neurophysiol 1999;110:1589±1600. Waberski TD, Gobbele R, Herrendorf G, Steinhoff BJ, Kolle R, Fuchs M, Paulus W, Buchner H. Source reconstruction of mesial±temporal epileptiform activity: comparison of inverse techniques. Epilepsia 2000;41(12):1574±1583. Wagner M. ekonstruktion neuronaler StroÈme aus bioelektrischen und biomagnetischen Messungen auf der aus MR-Bildern segmentierten Hirnrinde (doctoral thesis), Aachen: Shaker Verlag, 1998. Wagner M, Fuchs M, Wischmann HA, Ottenberg K, DoÈsel O. Cortex segmentation from 3D MR images for MEG reconstructions. In: Baumgartner C, editor. Biomagnetism: fundamental research and clinical applications, Amsterdam: Elsevier Science (IOS Press), 1995. pp. 433±438. Wang JZ, Williamson SJ, Kaufman L. Magnetic source images determined by a lead-®eld analysis: the unique minimum-norm least-squares estimation. IEEE Trans Biomed Eng 1992;39(7):665±675.
