JLE

Epileptic Disorders

MENU

Fear, anger and compulsive behavior during seizure: involvement of large scale fronto-temporal neural networks Volume 4, numéro 4, December 2002

Introduction

The semiology of seizures involving the limbic and paralimbic cortex may include intense affective changes. The best known examples are fear and anxiety, and there are generally considered to reflect the involvement of the amygdala during temporal lobe seizures [1]. Following these subjective affective changes, reported by the patient, more complex behavioral and emotional manifestations may appear during seizures such as intense screaming, dramatic changes in facial expression, gesticulations or complex motor activities. Complex motor phenomena occurring with epileptic seizures have been observed and described generally accompanied by disturbances of consciousness, but these have been rarely studied in correlation with electrophysiological data [2].These seizure-related, abnormal affective and gestural behaviours may involve some of the same processes as those underlying non-pathological behavior, but their pathophysiology remains largely unexplored. We report here, a patient in whom seizures initially involved the medial temporal region before involving the ventral prefrontal (orbito-frontal) cortex. Seizure semiology of the second part of the seizure included marked emotional disturbances (dominated by intense fear and anger) and a particular compulsive behavior. This patient underwent presurgical evaluation including, in particular, stereoencephalographic recordings (SEEG). Our purpose was to analyse the neurophysiological and anatomical correlates of these behavioural manifestations.

Methods

Subject

Presurgical evaluation was undertaken in a 40 year-old, right-handed woman (S.B.), who presented with drugresistant, partial seizures since the age of 12. S.B. was born after a normal full-term pregnancy. Psychomotor development was normal and there was no family history of epilepsy. Video-EEG monitoring of a number of seizures revealed a consistent pattern. Onset was characterized by arrest of activity with the preservation of consciousness and features suggesting mild anxiety. This anxiety progressed to intense fear with vegetative signs (mydriasis, rubefaction), and mild oro-alimentary automatisms. This initial period was followed about 15 s later by sudden agitation, facial expression expressing terror and anger, with screaming and biting usually of bedclothing or a handkerchief. The end of the seizure was marked by confusion and transient mood changes (sadness and anxiety as reported by the patient). After the seizures, S.B. confirmed having felt an intense fear and an irresistible, uncontrollable urge to bite into something. Ictal scalp EEG demonstrated rhythmic theta waves 5 s after the first clinical signs. This discharge appeared initially over the right temporal electrodes then spread to the whole right hemisphere and the anterior left region. MRI showed slight right hippocampal atrophy. Ictal SPECT was unavailable.

Stereoelectroencephalographic study

At the end of a non-invasive preoperative investigation, an SEEG exploration [3, 4] was performed. Its purpose was to determine the anatomical structures involved in the initiation and propagation of seizures, and the accurate limits for future cortical excision. It required the implantation of multiple leads (0.8 mm diameter, 5 to 15 contacts, 2 mm length, 1.5 mm apart), depth electrodes, which were introduced orthogonally through the usual double grid system attached to the Talairach stereotaxic frame, according to the non-invasive data obtained on the epileptogenic zone. The anatomical structures explored with depth electrodes in this patient are reported on Talairach’s proportionnal grid (figure 1).

Coherence study of compulsive symptoms during seizures

Our aim was to study the stereo-electroencephalographic correlates of seizures with particular respect to the intense fear and compulsive behavior. In addition to SEEG traces examination, we used coherence analysis of signals as a means of studying functional coupling between different regions of the brain. Coherent EEG activity has already been demonstrated to be associated with different cognitive processes in human and non-human mammals [5, 6]. We made the assumption that behavioral changes during seizures could be associated with changes in coherent activity of SEEG signals recorded from regions of the brain involved in the genesis of this symptomatology. In the present work, the coherence was computed using a bias-reduced estimator. Readers may refer to [7, 8] for detailed information on this estimator.

Statistical analysis of coherence values

Coherence values were compared before and during a given period of interest to study variations in the functional coupling of different cerebral structures. Two seizures were studied. Coherence values from different periods of interest were compared to identify the neural structures involved at the onset of seizure activity as well as during the emotional behavioral changes (EBC) (figure 2). To identify the structures initially involved, values obtained 15 s prior to seizure onset (Pre I) were compared with values obtained after seizure onset [I]. For the identification of the neural structures implicated in the intense emotional behavioral changes, values obtained during the ictal onset period [I] were compared with values obtained after the onset of EBC(EBC ). At each of the periods, interactions between regions of the brain containing intracerebrally-implanted electrodes were studied. The cerebral regions implicated in EBCwere identified as those regions for which coherence values were higher at EBCthan at pre-EBC. A methodological problem associated with studying coherence arises from the fact that coherence values are correlated in time (i.e. a sliding window is used). Coherence data cannot be used directly in statistical tests which evaluate differences in the distributions of two independent samples. In the present study, this problem was resolved by de-correlating the coherence series by means of an inverse autoregressive (AR) filtering operation. In brief, this operation is based on the assumption that a coherence series can be assimilated to a filtered noise with a normal distribution and that, as such, passing it through the inverse of its estimated AR filter results in normal noise distribution with zero mean and variance. After AR filtering, statistical tests of conformity may be performed on the two series of values. For inverse filtering, an AR model was estimated from the coherence series that were to be compared. Model order was determined from an Akaike criterion [9]. Model order 6 was often found to be appropriate. Resulting innovations were used in a classical F-test for significantly different variances. The F-test tests the hypothesis that two samples have different variances (i.e. that the null hypothesis is false). In this test, the constructed statistical F is the ratio of one variance to another such that ratios significantly less or greater than 1 will indicate significant differences. Small P values correspond to significantly different variances of the compared series and, by extension, to significantly different series of coherence values. For the present study, the level of significance was defined as P < 0.05.

Results

SEEG data

Interictal activity was abundant over the electrodes exploring the right temporal pole, amygdala and hippocampus. Two spontaneous seizures were recorded and one seizure was produced following electrical stimulation of the electrode exploring the anterior hippocampus. These seizures were strictly identical and reproduced the semiology recorded during previous video-scalp EEG recording. Seizures were marked by the emergence of a rythmic ictal discharge around 15 Hz over the internal leads exploring the temporal pole, the amygdala and the hippocampus (figure 3). At this stage, the patient complained of mild anxiety and vegetative changes (tachycadia, rubefaction), but was able to answer questions. Ten seconds later, a slower activity appeared over the leads exploring the frontal orbital region. After the start of these frontal changes, the behavior of the patient changed abruptly and the patient screamed, assumed a terrified facial expression and appeared to be compulsively biting the bedcloths.

Coherence study of compulsive behavior

We separated the results into two periods: the first corresponds to the period of initial ictal activity (period 2 in figure 2). Table I summarizes the coherence values obtained during these two periods between the SEEG signals of brain structures explored. A first network was identified and made up the epileptogenic zone. The structures involved were within the right anterior temporal regions (amygdala, temporal pole and hippocampus, temporal neocortex) (figure 4). The second period is the period in which the behavioral changes occurred. At this time and by comparison with the ictal period, there was a strong functional coupling between the amygdala, the orbito-frontal cortex and the frontal opercular region (F3), while a decrease in functional coupling between these regions and the dorsolateral region and the cingulate gyrus was apparent.

Discussion

This case provides a correlate between ictal emotional behavior and detailed electrophysiological data from depth electrode studies. The seizure semiology could be divided into two main periods. The first was mainly characterized by subjective symptoms, particularly a feeling of anxiety that gradually increased. The period following was dominated by a compulsive urge to bite something, in the context of intense fear, screaming, agitation, and dysautonomic changes. This part of the semiology was delayed, occurring 15 seconds into the right temporal lobe seizure. Indeed, SEEG recordings showed that seizures started in the mesial structures affecting the amygdala, the temporal pole and the hippocampus. The occurrence of compulsive behavior, screaming and intense fear coincided with the emergence of slow wave activity over the electrodes exploring the ventral pre-frontal region, suggesting a role for these structures in the changes in clinical symptomatology. Coherence analysis of SEEG signals was carried out as a means of studying the functional coupling of different regions of the brain. Coherence has already been used in depth electrodes temporal lobe epilepsy studies to define the neural structure interactions during seizures [8, 10, 11]. The synchronization between the different regions of interest was evaluated by an original coherence algorithm previously described [7, 8]. The two recorded spontaneous seizures shared similar features: ictal activity starting in the amygdala, the hippocampus and the temporal pole. Analysis of coherence in the first period demonstrated strong inter-relations between anterior temporal lobe structures. The neural network as described by coherence analysis, did indeed consist of the amygdala, the hippocampus, the temporal pole and the anterior temporal neocortex (middle temporal gyrus). In this initial period, the semiology characterized by dysautonomic changes and mild anxiety was probably related to the temporal limbic involvement. The amygdala is known to play a central role in the experience of fear in normal individuals [12], and anxiety or fear is the most common affective symptoms in TLEs [1]. In this patient, the seizures, the subsequent phase of intense emotional behavior with facial expressions suggestive of anger and terror, screaming, obvious vegetative changes and the compulsive biting were shown to be linked to a change in the neural networks involved. Indeed, a strong, significant increase in coherence value demonstrated that the epileptic network at this time consisted of the amygdala, the temporal pole, the prefrontal orbital and opercular regions, regions that are closely interconnected anatomically. Interestingly, these regions have already been shown to be involved in neural networks participating in emotion and cognition [13, 14]. In particular, the main function of the amygdala appears to be the linking of the perceptual representations to cognition and behavior on the basis of the emotional or social value of external stimuli [14]. The orbito-frontal region may serve to regulate and inhibit the activity of the amygdala, important roles in controlling impulsive, aggressive and violent behavior [15]. The epileptic discharges recorded during this patient’s seizures appear to markedly disturb the normal function of these structures, leading to the inappropriate manifestation of fear, screaming and impulsive biting. In addition, a decrease in coherence between the orbito-frontal cortex and the dorso-lateral pre-frontal cortex could enhance this phenomenon by reducing the inhibiting effect of these regions known to be involved in the conscious control of the emotions [16]. The decrease in the relationships with the anterior cingulate gyrus could also permit a "release" effect and thus the emergence of this particular behavior since the anterior cingulate cortex has been proposed as a key structure involved in translating intentions to actions [17]. It is also interesting to note that compulsive behaviors associated with obsessive-compulsive disorders (OCD) have been related to dysfunction in networks including the orbito-frontal cortex and sub-cortical structures (the ventral neostriatum, the globus pallidus and the thalamus) [18]. On the conceptual level, this case study demonstrates that partial seizures are organized according to neural networks that appear to share the structures of normal cognitive/emotional networks. This assumption of the common basis of both physiological and pathophysiological networks has been previously used to describe the organization of the epileptogenic zone of the temporal epilepsies [8] and to explain the memory disturbances occurring during certain seizures [19]. Definite conclusions cannot be arrived at on the basis of a single case, but this study provides a basis for further discussion. In particular, this type of study may allow us to better understand the basis of complex behaviors observed during seizures and may shed light on the neural networks underlying emotions and associated behavior in humans.

Acknowledgements: We thank Aileen Mc Gonigal for the revision of the english version of this paper.

Received March 25, 2002 Accepted August 5, 2002