Spatio-temproral EEG-MEG Analysis of Rolandic Spikes

G.J.M. Huiskamp¹, W. van der Meij,¹A.C.van Huffelen¹and O. van Nieuwenhuizen²
¹Department of Clinical Neurophysiology, ²Department of Child Neurology,
University Medical Center Utrecht, Heidelberglaan 100, 3584 CX Utrecht, The Netherlands

Abstract: Spatio-temporal aspects of the sources of spikes in benign rolandic epilepsy were investigated. Spike activity in 5 children all having general and/or focal seizure were recorded using simultaneous 64 channel EEG and 151 channel MEG. In a realistic volume conductor model a spatio-temporal MUSIC analysis was performed. Results show that in some cases sources well separated in space and time exist, in other cases only evidence for a single source is found.

Introduction

There have been indications that the source of Rolandic spikes cannot adequately be modeled by a single dipole. In a previous study careful analysis has shown that in some cases the source could be characterized by two dipolar components with a different orientation, whereas the difference in position could not be resolved [1]. It was demonstrated that the occurrence of such a double spike has a significantly stronger association with the presence of epilepsy than the single spike has. These studies were based on 32 channel EEG measurements, and dipole modeling was done using a three shell spherical model. However, in recent simultaneous EEG and MEG study only spikes that could be characterized by a single dipolar source were found [2].

In this paper we present results for a group of 5 patients with rolandic epilepsy for which both a high resolution EEG and MEG was available. The main focus was on whether a single or more complex source was present and, in the latter case, what the relative position and timing of the different components of the source activity was. In order to avoid any a priori choice of the number of independent sources a MUSIC analysis was performed on a moving time window encompassing the spikes.

Methods

Five patients (A,B,C,D,E) were considered (3 female, 2 male, age between 5 and 12). All patients had rolandic epilepsy with focal and/or generalized seizures. 64 channel EEGs were registered at 625 Hz, simultaneous with 151 channel MEGs (Omega 151, CTF systems Inc.) In cases where the simultaneous EEG measurement was not possible, an 84 channel EEG at a sampling rate of 1024 Hz was measured separately (BioSemi Mark-6, Brainstar4.0). Electrode positions and the position of 3 anatomical marker locations (nasion and left and right pre-auricular points) were measured using a magnetic tracking device (Polhemus ISOTRAK II).

3D T1 MRI images were acquired on a Philips ACS-NT 1.5 T-scanner at a resolution of 0.89 x 0.89 x 1.5 mm. At the location of the three anatomical landmark positions that were also marked in the EEG and MEG measurements vitamin E capsules were applied that could be identified accurately in the registration. These were used to geometrically match the data-sets involved. Head-shape information present in the sampled electrode positions was used as well.

Clear spikes having the same morphology and spatial distribution were identified in the data by an experienced neurophysiologist. Template matching was applied in order to obtain a population of synchronized, highly similar spikes. These spikes were then averaged in order to reduce background activity.

The anatomical MRI data were processed using CURRY3.0 (NEUROSCAN). Segmentation of the skin and the outer and inner surface of the skull were generated. From these segmentations triangulations, containing typically 1500 points each, were made. Based on these triangulations BEM models were created. The skull-to-skin conductivity ratio used in this study was 1/20 [3].

For each of the sets a solution-domain surface was constructed by deflating the inner skull surface. The depth of the surface was chosen in such a way that a maximal cross section of the rolandic area was obtained. The solution surface consisted of typically 1500 points, with an average inter-point distance of 5 mm in the rolandic area.

A moving time window of approximately 20 ms (11 samples for 625 Hz data) was defined for which a "classical" MUSIC analysis for rotating dipoles defined in the nodal points of the solution surface was performed [4]. The width of the window was inspired by previous work on the rolandic double spike [1] and chosen to be small enough to enable the capture of possible sequential events separately (cf. a moving dipole analysis) but wide enough to allow for a signal/noise-space separation at a dimension larger than one (cf. multiple dipoles). This separation was defined by the rank at which 90% of the signal in the window was explained, with a minimum value for this rank of 2.

The data was analysed by extracting the time instance of first occurrence of a relative maximum of the MUSIC metric, the location of this relative maximum when it reaches its peak value, the time elapsed till the occurrence of a second relative maximum, and the location of this relative maximum when it reaches its peak value.

Figure 1. MUSIC metric plotted over rendering of matched cortex geometry at three time instances for 64 channel EEG of patient A.

Results

For all patients the time differences (in ms) between first occurrence of the first and second source, the distances between the distribution maxima (in cm) and the dominating direction for this distance are summarized in Table I.

A detailed plot of results for the 64 channel EEG for patient BA is shown in figure 1. The MUSIC metric distribution for different time instances is shown here plotted over a rendering of the matched cortex geometry.

DISCUSSION

Results in the literature show that the rolandic spike can in some cases be modeled by a single current dipole, in other cases a more complex source is indicated. When using single or multiple dipole models the correct a priori choice on the number of dipoles is important: both under-and overestimation of this number lead to incorrect results. Although in the MUSIC approach an estimate of the dimension of signal space has to be made, an over-estimation of that dimension does not lead to significantly different results.

Previous studies from our group have shown that single rolandic spikes in about 50% coincided with clinical epilepsy. Double spikes have a more than 90% association with clinical seizures. In the present study all patient considered had rolandic epilepsy with focal and/or generalized seizures. Results show that the characterization of the sources of the spikes observed in these patients can vary from definitely multiple, separated well in both space and time (patient A), to almost certainly single (E).

The results in figure 1 suggest that an initial source is present at the arm/hand area, and that some 10 ms later a second source much lower, in the mouth area starts building up. This source proceeds even lower along the central sulcus later. A similar result was obtained for other patients. In the literature on benign rolandic epilepsy the existence of spikes evoked by electrical stimulation of the median nerve has been reported on [5]. It is tempting to suggest that in those cases a pathway similar to the one present in the spontaneous data is followed.

TABLE I
Summary of results. Time difference in ms, distances in cm and dominant direction of displacement
between first and second maximum for MEG and simultaneous or separate 84 channel EEG.
Results in italics could only be approximated

	MEG	sim EEG	EEG (84)
A	21 ms 1.2 cm -z	4.0 cm - z	-
B	30 ms 2.3 cm -z	-	22 ms 2.5 cm -z
C	19 ms 1.4 cm -y	0.5 cm -y	-
D	? ?	27 ms 2.6 cm -z	-
E	22 ms ?	-	27 ms 1.0 cm -z

References

[1] W. van der Meij, G.H. Wieneke and A.C. van Huffelen, "Dipole source analysis of Rolandic spikes in benign rolandic epilepsy and other clinical syndromes, Brain Topogr vol. 5, pp. 203-213, 1993

[2] E. Pataraia, M.Feucht, G. Lindinger et al., "Combined EEG and MEG study of rolandic discharges in benign childhood epilepsy", in BIOMAG2000, Proc. 12th Int. Conf. on Biomagnetism Espoo., 2001, pp. 447-450.

[3] T.F. Oostendorp, J. Delbeke and D.F. Stegeman DF, "The conductivity of the human skull: results of in vivo and in vitro measurements", IEEE Trans Biomed Eng , vol. 46, pp. 1487-1492, 2000.

[4] J.C. Mosher, P.S. Lewis and R.M. Leahy, "Multiple Dipole Modeling and Localization from Spatio-Temporal MEG Data", IEEE Trans Biomed Eng , vol. 39, pp. 541-557, 1992.

[5] P. Manganotti and G. Zanette, "Contribution of Motor Cortex in Generation of Evoked Spikes in Patients with Benign Rolandic Epilepsy", Clinical Neurophysiol, vol. 111, pp. 964-974, 2000.