Linking generalized spike-and-wave discharges and resting state brain activity by

Abstract

Purpose: To illustrate a functional interpretation of blood oxygen level-dependent (BOLD) signal changes associated with generalized spike and wave discharges in patients with absence seizures and to demonstrate the reproducibility of these findings in one case.

Methods: In a 47 year-old patient with frequent absence seizures BOLD signal changes during generalized spike and wave discharges (GSWD) were mapped using simultaneous and continuous electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) at 1.5 T and 6 months later at 3 T. GSWD were modelled as individual events and as blocks.

Results: The patient studied exhibited frequent generalized spike wave activity with temporal properties ideal for study with EEG-fMRI. Highly reproducible GSWD associated fMRI signal decreases (‘deactivations’) were seen in bilateral frontal and temporo-parietal cortices and the precuneus, in addition to activations in occipital cortex and, at 3 T, the posterior

thalamus.

Discussion: The GSWD associated changes seen here involve cortical regions that have been shown to be more active at conscious rest compared to sleep and to various types of

extroverted perception and action. These regions have been proposed to constitute the core of a functional ‘default mode’ system. We propose that the findings of 'deactivation' of this distributed brain system during GSWD mirrors the clinical manifestation of GSWD, i.e.

absence seizures. Furthermore, we suggest that these deactivations may reflect the functional consequences of GSWD on physiological brain activity at rest rather than direct hemodynamic correlates of epileptic discharges.

Introduction:

3 Own contributions: study design, data acquisition and analysis, result interpretation, entire manuscript preparation

When combined with simultaneous EEG recording, BOLD fMRI techniques permit the

identification of brain regions affected by GSWD (Aghakhani et al., 2004; Archer et al., 2003a;

Baudewig et al., 2001; Salek-Haddadi et al., 2003b; Warach et al., 1996b). Ictal symptoms are not necessarily ‘positive’, such as movements or sensations but can be ‘negative’, e.g.

absences. One question is whether such negative symptoms are associated with

‘deactivation’ as indexed by hemodynamic signals. Another current question concerns the topographical relation between electrical and hemodynamic brain signals. If for instance epileptic activity on surface EEG is not focal but generalized, would one predict that activation or deactivation occur reproducibly and globally throughout the brain?

Methods Patient

We studied a 47 year-old male (written informed consent, ethics committee approval) with juvenile absence epilepsy, onset at age 8, with frequent absence seizures and less than one generalized tonic-clonic seizure per year. Intellectual development, neurological examination and structural imaging findings were all normal. At the time of investigation, more than 10 typical absence seizures per hour were seen on clinical observation as a sudden cessation of activity but without additional motor symptoms. EEG showed a normal background

interrupted by frequent 2-3/s GSWD lasting between <1 and 20 seconds, and occasionally up to 40 seconds. Absence seizures had proven refractory to treatment with phenytoin,

valproate, phenobarbitone, clonazepam, ethosuximide and primidone. The patient was taking 750 mg of Primidone at the time of both experiments. He was resistant to further changes in medication.

fMRI data acquisition

The patient was studied at 1.5 T (experiment I) and, for reproducibility at follow-up 6 months later at 3 T (experiment II). Two 20 minutes imaging sessions were acquired at each

experiment. During scanning the instruction was to lie still with eyes closed (“rest”). Imaging parameters: T2*-weighted echo-planar images covering the entire cerebrum; 1.5 T Siemens Vision (Erlangen, Germany), voxel size 3.44 x 3.44 x 4 mm, 19 slices with 1 mm gap in 2.8 s, 300 volumes, TR/TE 4000/50; 3 T Siemens Trio (Erlangen), voxel size 3 x 3 x 4 mm, 20 slices with 1 mm gap in 1.3 sec, 300 volumes, TR/TE 2666/20, vacuum head cushion.

EEG and artifact reduction

EEG was recorded using the BrainAmp MR EEG amplifier (5 kHz), Brain Vision Recorder software (Brainproducts, Munich, Germany) and the BrainCap electrode cap (Falk Minow Services, Herrsching-Breitbrunn, Germany) with 29 electrodes (10-20 system, reference between Fz and Cz). Off-line EEG artifact subtraction was performed, as implemented in the Brain Vision Analyzer (Brainproducts) (Laufs et al., 2003a).

Data analysis

We used statistical parametric mapping (SPM2 [http://www.fil.ion.ucl.ac.uk/spm]) for image realignment, normalization, spatial (10 mm full width at half maximum Gaussian kernel) and temporal (128 s high pass filter) smoothing and statistical analysis. Epochs of GSWD were visually identified and used to define a boxcar which was convolved with the hemodynamic response function (HRF; block design). In addition, a template matching algorithm (based on amplitude and correlation measures, Ingmar Gutberlet, VisionAnalyzer, Brain Products) was used to individually mark the spikes of GSWD. These markers were treated as events and convolved with the HRF and used as a regressor for an event-related fMRI analysis. In a third model we grouped runs of GSWD into two event types, depending on whether they lasted for longer or less than 1 second to test for differences in the fMRI response to short and long runs. Motion parameters obtained from realignment were always included as confounds.

An eigenimage analysis was undertaken and is presented as supporting material.

Results

In all 4 sessions frequent GSWD were seen (figure A&B) and the patient was in ‘quiet rest’ as monitored by the EEG showing a normal awake background with continuous 9 Hz posterior alpha rhythm. There were no drops in vigilance as assessed by looking for vigilance/sleep criteria on the EEG (alpha amplitude variation, slowing, eye movements, sleep grapho-elements, heart rate). Reproducible negative correlations with GSWD were detected

bilaterally in dorsal prefrontal and temporo-parietal cortices and the precuneus and posterior cingulum (figure C) with positive correlations in the occipital cortex and bilateral inferior parietal areas. At 3 T, additional significant activations were found in the posterior thalamus (figure D). The block and the event related design showed the same signal changes. The deactivation patterns from short and long runs of GSWD were similar to the above analyses and to one another (significance was lower for short runs as they were less in number).

Eigenimage analysis confirmed high reproducibility of the findings and demonstrated

functional connectivity of the areas found to deactivate during GSWD (see supporting material).

Discussion

In accordance with our findings, previous fMRI studies have reported predominantly negative cortical BOLD responses or ‘deactivation’ in association with GSWD in patients with

idiopathic generalized epilepsy (IGE, table) (Aghakhani et al., 2004; Archer et al., 2003a;

Salek-Haddadi et al., 2003b). Yet, an interpretation of the functional significance of deactivations during GSWD remains lacking, in particular of those regions reported here.

Meta-analyses of studies in healthy subjects have established that across different functional activation paradigms a similar set of brain areas will consistently display deactivations

(Mazoyer et al., 2001; Raichle et al., 2001a): activity in bilateral medial frontal and parietal areas and the precuneus is spontaneously higher when the brain is at ‘rest’ than when the brain is engaged in extroverted perception and action. Activity in this functional network has thus been termed a “default mode” of brain function, when cognitive processes revolve about the self, its past and its future (Raichle et al., 2001a).

This introspective cognitive state can be regarded a cornerstone of consciousness. The pattern of cortical deactivation we observed in these “default mode” areas (figure C) has also been identified in slow wave sleep compared to wakefulness (Maquet, 2000); and the precuneus has been reported to reflect the difference between wakefulness and clinical states where consciousness is impaired as in coma and anaesthesia (Laureys et al., 2004a).

We could recently link the “default mode“ areas to another EEG feature, namely activity (power) in the range of 17-23 Hz during conscious rest and also show their connectivity (Laufs et al., 2003b). Here, in the context of GSWD, we also found these areas to be

functionally connected (see supporting material). We suggest that the regions of BOLD signal decreases indicate areas where GSWD are associated with a disturbance or suspension of this physiological conscious rest. The particular deactivations might not be specific for or causally related to GSWD per se. Instead, they may provide an fMRI signature of the negative clinical phenomenology of absence seizures similar to sleep, anaesthesia or coma - impaired consciousness, dynamically occurring even on a second-by-second basis.

We hypothesize that whilst our patient was at rest and not experiencing absence seizures, the ‘default mode’ brain areas were active as part of normal processing during conscious rest. This conscious rest (with normal background EEG activity) was interrupted by GSWD, presenting in fMRI as a relative deactivation of that mode. Clinical absences during

prolonged runs of GSWD were evident outside the scanner. Nevertheless, the same pattern of deactivation was found for short and long runs of GSWD. We suggest that in our patient even short discharges might lead to a (subclinical) interruption of cognitive processes (as reported by others (Aarts et al., 1984)). However, any experimental interference to confirm this assumption would have obliterated the resting state. For example, a simple button press-task to monitor the patient’s state of consciousness in the scanner would have suspended the default mode in the first place, and during GSWD one might, for instance, have seen deactivations in the [motor] areas recruited by that task.

Apart from the common pattern of deactivation discussed here, there is variability in the findings of EEG-fMRI studies of IGE (Aghakhani et al., 2004; Archer et al., 2003a). This can be explained by the fact that different syndromes of IGE have been studied and the clinical manifestation of different types of absences is not uniform (Gloor, 1986). Reflecting these clinical differences, varying brain areas may be influenced by GSWD and may explain the different activations seen, including the occipital activation in this patient (figure D). An alternative account of this variability, however, is that it expresses the physiological (and between session) variability of resting state brain activity rather than differences between absence syndromes. In this context, the fMRI activations represent brain areas involved in the mental activity disrupted by GSWD. Combining cognitive paradigms with EEG-fMRI experiments might lead to a more realistic characterization of how epileptic EEG activity interferes with the patients’ cognitive function.

The presence of thalamic activation seen here at 3 T - but not at 1.5 T - is in keeping with higher sensitivity of fMRI at 3 T and supports the involvement of the thalamus in generating GSWD. A reciprocal relation between thalamo-reticular and cortical structures in generalized epilepsies has been supported by findings in animal models, electrophysiological and

functional imaging experiments (Aghakhani et al., 2004; Salek-Haddadi et al., 2003b;

Timofeev and Steriade, 2004). Studies using deep brain stimulation and PET have

demonstrated an inhibitory input to the thalamus being associated with an increase in thalamic but a decrease in frontal, parietal and temporal blood flow, probably caused by inhibiting thalamo-cortical connections (Hershey et al., 2003).

As one potential explanation we suggest on the basis of our observations and that of others (table) that functional changes in the ‘default mode’ network reproducibly occur during GSWD and might be related to the clinical manifestation of absence seizures.

Supporting Material Method

Singular-value decomposition (principal component analysis) of the fMRI time series was performed by using the multivariate linear models toolbox as implemented in SPM99 (Kherif et al., 2002) after removing the session mean and the variance explained by the realignment parameters. The correlation coefficients between the computed eigenvariates

and the smoothed, filtered, and hrf-convolved regressor modeling each GSWD as a single event were calculated by using MATLAB (Mathworks, Inc., Sherborn, MA, U.S.A.).

Results and discussion

In every session, the second BOLD-eigenvariate showed the maximum correlation with the GSWD regressor, ranging from –0.7 to –0.8, the negative values reflecting their inverse relation. The related eigenimages very closely matched the topographic “deactivation”

pattern that we determined (Fig. 1D), whose appearance in one single eigenimage suggests functional connectivity, as demonstrated previously in this network. The high variance

of 9% to 14% of the variance in the fMRI dataset explained by this spatial mode points to the prominence of activity in these areas during conscious rest, which is then suspended.