Scalp electroencephalography (EEG) is an important tool for the investigation of patients with epilepsy. It can help to diagnose and classify epilepsy and localize sources of epileptic activity. However, EEG is characterized by a number of disadvantages. The EEG has low spatial resolution with a limited ability to detect epileptic activity from deep brain structures and is characterized by the inverse problem with ambiguous source solutions for a given spatiotemporal distribution of a cerebral current (Michel et al., 2004). In contrast, functional magnetic resonance imaging (fMRI) is characterized by high spatial resolution, is equally sensitive to signals of deep and superficial brain structures, and represents a measurement of hemodynamic changes in the brain. By combining EEG with fMRI it is possible to detect blood oxygenation level–dependent (BOLD) signal changes associated with ictal and interictal epileptiform discharges (IEDs) obtained from the scalp EEG, even if deep brain structures are involved in their generation. In the last 15 years, EEG-correlated fMRI studies have been used to delineate epileptic networks noninvasively. Most of the studies were carried out in adult patients with epilepsy (for recent comprehensive review see Gotman & Pittau, 2011). However, childhood epilepsies differ from adult epilepsies regarding etiology, pathogenesis, seizure semiology, EEG patterns, and prognosis. The immature brain is more prone to develop seizures, and epileptic discharges are more frequent and less localized in children than in adults. Furthermore, many epilepsy syndromes are age-specific and manifest within only a certain age range (Ben-Ari, 2006). It can be suggested that the neuronal networks underlying IEDs in children may be influenced by brain development and differ from those in adults. Moreover, some epileptic syndromes and encephalopathies can only or predominantly be investigated in children, such as, for example, West syndrome, benign epilepsy with centro-temporal spikes (BECTS), or Dravet syndrome. Hence, EEG-fMRI recordings in children may contribute to a better understanding of pathophysiologic mechanisms of epilepsy in general. This review summarized EEG-fMRI studies that have been performed in children with epilepsy.
By combining electroencephalography (EEG) with functional magnetic resonance imaging (fMRI) it is possible to describe blood oxygenation level–dependent (BOLD) signal changes related to EEG patterns. This way, EEG-pattern–associated networks of hemodynamic changes can be detected anywhere in the brain with good spatial resolution. This review summarizes EEG-fMRI studies that have been performed in children with epilepsy. EEG-fMRI studies in focal epilepsy (structural and nonlesional cases, benign epilepsy with centrotemporal spikes), generalized epilepsy (especially absence epilepsy), and epileptic encephalopathies (West syndrome, Lennox-Gastaut syndrome, continuous spike and waves during slow sleep, and Dravet syndrome) are presented. Although EEG-fMRI was applied mainly to localize the region presumably generating focal interictal discharges in focal epilepsies, EEG-fMRI identified underlying networks in patients with generalized epilepsies and thereby contributed to a better understanding of these epilepsies. In epileptic encephalopathies a specific fingerprint of hemodynamic changes associated with the particular syndrome was detected. The value of the EEG-fMRI technique for diagnosis and investigation of pathogenetic mechanisms of different forms of epilepsy is discussed.
Methodologic Aspects of EEG-fMRI Studies in a Pediatric Population
Figure 1 summarizes different steps of an EEG-fMRI investigation. In all EEG-fMRI studies performed in children, the same equipment, the same algorithms for correction of gradient and ballistocardiogram artifacts, and the same statistical analyses were used as in adults. In summary, after correction for gradient and ballistocardiogram artifacts, IEDs are marked in the EEG studies obtained inside the MR scanner and taken as events. As an alternative approach, the IED-associated voltage map can be used to create a time series of events (Grouiller et al., 2011; Elshoff et al., 2012). In a standard analysis, the timing of events is convolved with the canonical hemodynamic response function (HRF), which shows a peak approximately 5 s after the event. Statistical maps are computed showing voxels that correlate significantly with the marked events in the EEG (Friston et al., 1995). However, a standard HRF does not take into account that shape and latency of the HRF might vary with age, different brain regions, or altered response in epilepsy. Several studies have shown that the sensitivity of EEG-fMRI studies can be improved if a more flexible and individually adjusted HRF is applied. The individual adjustment of the HRF can be achieved by including the time and dispersion derivative of the HRF, in addition to the standard one (Hamandi et al., 2006), by applying a set of four HRFs with peaks at 3, 5, 7, and 9 s (Bagshaw et al., 2004) or by estimating noncanonical HRFs (van Houdt et al., 2010).
EEG-fMRI recordings in children are challenging due to a lack of cooperation, especially if the children are young or mentally disabled, which can often be the case with epilepsy patients. For this reason, it is necessary to minimize anxiety, so that the child will agree to enter the scanner and remain inside the scanner for the duration of the recording, and to ensure that the child stays motionless throughout the entire examination (30 min on average). In some studies, sedation is needed for children undergoing EEG-fMRI recordings (Jacobs et al., 2007; Siniatchkin et al., 2007a; Jacobs et al., 2008a,b; Moeller et al., 2008a; Siniatchkin et al., 2010, 2011). In our laboratory, most children were sedated using chloral hydrate (50 mg/kg body weight). Chloral hydrate is a well-tolerated sedative that only minimally influences epileptic activity and is often used for sleep-EEG recordings in specialized epilepsy centers (Thoresen et al., 1997). However, sedation has been shown to influence hemodynamic parameters and fMRI results in activation studies (Born et al., 1998). By comparing hemodynamic changes in epilepsy patients with and without sedation, we were unable to describe significant effects of chloral hydrate on the results of EEG-fMRI (see Jacobs et al., 2007; Moeller et al., 2008a; Moehring et al., 2011). In nonsedated children (usually older and without mental retardation), the anxiety can be minimized, if parents or someone familiar to the child stays in the MR room during data acquisition; if EEG-fMRI is simulated first to allow the child to get used to conditions of the test; if external stimulation (video and audio tapes) is used before the test to distract the child; or if the child is exposed to the scanner surroundings in a surrogate scanner (Poldrack et al., 2002). However, even well-prepared children move extensively (head and body movements) during data acquisition. To reduce the effect of head motion on EEG-fMRI results, different algorithms have been developed to improve quality and sensitivity of fMRI analysis (Wilke, 2012). Still, if pronounced movements occur, the data sets have to be discarded.
Modeling IED-related BOLD changes in children
Even if the datasets are of sufficient quality, different, age-specific aspects of the analysis have to be taken into account to increase sensitivity of EEG-fMRI studies in children. For example, the HRF to external stimuli is known to change over the course of normal development, most notably during infancy (Richter & Richter, 2003). Whereas acoustic and visual stimulation results in positive BOLD responses (activations) in adults, several studies have reported negative BOLD responses (deactivations) to the same sensory stimulation in infants (for review see Poldrack et al., 2002). Jacobs et al. underscored the impact of age on the spike-related BOLD signal changes and demonstrated that children with focal lesional epilepsy show deactivations more frequently than activations in the epileptogenic zone compared to adults (Jacobs et al., 2007). The shape of the HRF was different, especially in young children, where it peaked significantly later than the typical HRF in older children and adults (Jacobs et al., 2008a). Modeling spike-related BOLD signal changes in children might be even more complex, since significant activations in some children may already occur several seconds before the IEDs are detected on the scalp EEG (Moeller et al., 2008a, 2009a; Jacobs et al., 2009). The BOLD signal changes preceding IEDs were better localized than later BOLD signal changes (Jacobs et al., 2009). It can be suggested that the preceding BOLD signal changes represent synchronized discharges detectable with depth electrodes but not visible on the scalp EEG (Pittau et al., 2011). Whatever the explanation, if the early activation is not considered, the sensitivity of EEG-fMRI studies may suffer.
Finally, changes in vigilance (drowsiness and sleep) and the effect of antiepileptic medication itself probably modulate the BOLD signal and, hence, influence the fMRI results (Born et al., 1998). For instance, more negative BOLD responses were observed if children slept during the EEG-fMRI studies (Moehring et al., 2008). A recent study showed that modeling sleep-specific activity in children with epilepsy could increase the sensitivity of IED-associated BOLD signal detection (Moehring et al., 2011). Systematic studies are needed to describe the effect of sleep and medication in more detail. However, in primary generalized paroxysms, age, sleep, sedation, and antiepileptic medication did not seem to have an influence on network activation (Moeller et al., 2008a).
Idiopathic Focal Epilepsies
Idiopathic focal epilepsies represent a group of epileptic disorders that is related to well-localized focal epileptic EEG patterns, which are under strong developmental influence (Dalla Bernardina et al., 2005). Early applications of EEG-fMRI in children focused on benign epilepsy with centrotemporal spikes (BECTS) or rolandic epilepsy. Seizures in BECTS typically begin with paresthesia and jerking of the mouth, face, and hand, usually with a preserved level of consciousness, a fact that indicates an origin of epileptic activity in the sensorimotor cortex. Accordingly, EEG-fMRI studies in patients with BECTS have shown IED-associated BOLD signal changes in the sensorimotor cortex, whereas distant BOLD signal changes were interpreted as propagated activity (Archer et al., 2003; Boor et al., 2003, 2007; Lengler et al., 2007; Siniatchkin et al., 2007b). Note that even in patients with well localized IEDs, electrical source analysis was needed to differentiate between BOLD signal changes related to the initiation of epileptic activity from BOLD effects that can be attributed to propagated activity (Boor et al., 2007). The BOLD response to focal IED in BECTS was different from the canonical shape of the HRF, underscoring the importance of more variable HRF in pediatric studies (Masterton et al., 2010). In other idiopathic focal epilepsies (benign occipital lobe epilepsies of a Gastaut type and Panayiotopoulos type), Leal et al. (2006, 2007) found an activation pattern in different cortical occipital and parietal areas that corresponded well with the localization of IED. The authors suggested that EEG-fMRI provides a more satisfactory mapping of the irritative zone than that obtained from EEG source analysis. In summary, EEG-fMRI studies in children with idiopathic focal epilepsies were able to describe positive BOLD signal changes that corresponded well with brain areas that play an essential role in the pathogenesis of these epilepsies.
Symptomatic and Cryptogenic Focal Epilepsies
The development of EEG-fMRI was clinically motivated to localize regions that were presumed to generate focal IEDs. The majority of EEG-fMRI studies were performed in adult patients with focal epilepsy. Many of these studies have demonstrated a good concordance between BOLD signal changes and electroclinical data (recent studies Thornton et al., 2011; Pittau et al., 2012). Although previous studies compared EEG-fMRI results and intracranial recordings that were recorded on different days, it is now possible to perform an EEG-fMRI recording in patients with intracranial electrodes. Such studies show BOLD responses close to the electrodes from which spikes were recorded (Vulliemoz et al., 2011; Cunningham et al., 2012). Methodologic refinement, for example, the use of multiple HRFs or the fitting of individual HRFs and scanning at 3 Tesla rather than at 1.5 Tesla, increase the sensitivity of BOLD signal detection (Gholipour et al., 2011; Pittau et al., 2012). Few studies have evaluated the use of EEG-fMRI in presurgical investigation. A study in patients with medically intractable epilepsy suggested that EEG-fMRI may add useful information in the preoperative workup (Zijlmans et al., 2007). In patients with nonlesional frontal lobe epilepsy, a concordance between positive BOLD signal changes and postoperative pathologic analysis or other imaging modalities was found (Moeller et al., 2009b). An example of the comparison between EEG-fMRI results and methods of the presurgical evaluation in a child with focal epilepsy is shown in Figure 2. A postsurgical study showed that a surgical removal, which included the areas of positive BOLD response, was associated with a good postsurgical outcome (Thornton et al., 2010, 2011). Although some of the above-mentioned studies have included a few pediatric patients, only few studies have investigated specific groups of children with focal epilepsy.
De Tiege et al. (2007) presented for the first time results of EEG-fMRI studies in six children with symptomatic and cryptogenic pharmacoresistant focal epilepsy. In four children (66%), activations colocalized with the presumed location of the epileptic focus, and colocalization was seen for both activation and deactivation in another patient. Moreover, EEG-fMRI results were concordant with either invasive EEG recording (one patient), with brain lesion (two patients), or with ictal single-photon emission computed tomography (SPECT; two patients) which suggested a potential role of this method in children to noninvasively map hemodynamic changes associated with epileptic activity and delineate the epileptogenic zone.
Jacobs et al. (2007) analyzed 13 children with pharmacoresistant lesional focal epilepsy. In 84% of the cases, BOLD responses were localized in the lesion or presumed irritative zone. In contrast to studies in adults (Salek-Haddadi et al., 2006), deactivations in the lesion and the irritative zone were more common than activations. In another study of Jacobs et al. (2008b), five children with tuberous sclerosis complex (TSC) and pharmacoresistant focal epilepsy were studied. A BOLD response was found in at least one tuber localized in the lobe responsible for spike generation and presumed seizure onset zone (according to EEG-video monitoring) in all patients. In four patients, the same tubers were involved in generation of topographically different spikes, and the BOLD changes were always multifocal and sometimes involved tubers that were distant from the spike field. Another study of Jacobs et al. (2008b) demonstrated extended epileptogenic networks in patients with TSC that were more expanded than networks described in positron emission tomography (PET) and SPECT studies (Chugani et al., 1998).
In a group of six children with unambiguous focus localization (validated through ictal EEG, PET, and ictal SPECT) the hemodynamic changes corresponded to the epileptogenic zone in four children. Combined source analysis in these patients suggested that distant BOLD signal changes might be explained by propagated epileptic activity (Groening et al., 2009). The study of Groening et al. (2009) suggested that electrical source analysis has a better sensitivity in localizing the epileptogenic zone in children than EEG-fMRI. Moreover, the source analysis revealed clear advantages in separating the brain areas of the initial epileptic activity from areas involved in the propagated activity. The advantages of the source analysis were illustrated in a recent study by Elshoff et al. (2012). In this study, results of the electrical source analysis were compared to results of EEG-fMRI recordings in nine patients with pharmacoresistent focal epilepsy and who underwent epileptic surgery with favorable outcome (Engel class I and IIb). Although the results of the source analysis were concordant with the resection area in all patients, EEG-fMRI revealed areas of activation within the resection area in only four cases. A case report in a 2-year-old patient with hypothalamic hamartoma suggests that seizure propagation pathways can be investigated by dynamic causal modelling. The authors discussed that the knowledge of propagation pathways might contribute to decision making in the presurgical evaluation (Murta et al., 2012).
In summary, EEG-fMRI studies may represent a valuable option in the localization of the epileptogenic zone in some patients with pharmacoresistant focal epilepsies. However, the sensitivity of EEG-fMRI studies in children is lower than in adults, and a correspondence between the areas of activation and the epileptogenic zone can be found in only some of the children. At the moment, the EEG-fMRI technique only represents one option among others (e.g., EEG source reconstruction, PET, SPECT) in the presurgical setup. Further methodologic and validation studies are needed to evaluate and improve the impact of the technique for children with focal epilepsies.
EEG-fMRI in Idiopathic Generalized Epilepsy
Idiopathic generalized epilepsy (IGE) is characterized by an EEG pattern with generalized spike wave discharges (GSWDs) typically arising from normal background activity. Early EEG-fMRI studies investigated adult patients with idiopathic generalized epilepsy who showed short GSWD paroxysms in the EEG (Aghakhani et al., 2004; Hamandi et al., 2006). These studies confirmed that the thalamus is activated during GSWDs, but also showed a deactivation in the medial frontal cortex, precuneus, lateral parietal, and frontal cortex, that is, areas of the so-called default mode network (DMN). It has been proposed that the DMN represents a physiologic baseline for processes of attention and working memory and supports dynamic integration of cognitive and emotional processing (Raichle & Mintun, 2006). It has been suggested that disruption of the resting state activity by pathologic processes (e.g., those that give rise to spikes) may be related to alterations in cognitive function and be a possible mechanism that may underlie cognitive deficits in epilepsy (Gotman et al., 2005). Although most studies on IGE were performed in adult patients with long-lasting medically treated epilepsy, Moeller et al. (2008b) investigated absence seizures in drug-naive children with newly diagnosed epilepsy and confirmed thalamic activation along with deactivation in the DMN areas and the caudate nucleus. In a group analysis, Carney et al. (2010) showed that the brainstem is also involved during absence seizures. Subgroups of absence patterns with either predominantly positive or negative BOLD signal changes in the dorsolateral prefrontal cortex seem to exist and might have phenotypic and genetic implications (Carney et al., 2012). But how are absences initiated? Animal studies in genetic models of absence epilepsy strongly suggest that GSWDs are triggered in a restricted region of the somatosensory cortex (Meeren et al., 2002). A sliding window analysis of human absences showed that areas of the DMN and the caudate nucleus were involved significantly earlier than the thalamus. Early patient-specific BOLD signal changes may mirror a cortical focus (Moeller et al., 2010a). An example of a sliding window analysis of absences can be found in Fig. 3. In a dynamic group analysis of absence seizure, Bai et al. (2010) showed that BOLD signal changes in the orbitofrontal cortex preceded the onset of absence seizures by up to 14 s, whereas negative BOLD signal changes in default mode areas were seen up to 20 s after the end of absence seizures. This analysis indicated that there are long-lasting BOLD signal changes that are not detectable by conventional HRF modeling in many regions.
There are inconsistent results regarding the question of whether BOLD signal changes might occur prior to GSWDs: whereas BOLD signal changes preceding GSWDs have been reported in some adult patients with GSWDs, in children with polyspike wave paroxysms, and a group of patients with absence seizures (Moeller et al., 2008a; Bai et al., 2010), no preceding BOLD signal changes have been detected in other studies (Moeller et al., 2008b, 2011a). Early parietal BOLD responses prior to absences support the hypothesis that changes in activity within the default mode areas could facilitate the occurrence of GSWDs (Vaudano et al., 2009; Carney et al., 2010). Further studies are needed to investigate preceding hemodynamic changes before GSWDs are seen on the scalp EEG in detail. Now, it can be suggested that the conflicting results may be attributed to differences in the analysis or different samples of patients.
Why do patients show cognitive impairment during absences? EEG-fMRI studies with simultaneous testing during absences suggest that absences that are associated with a stronger cognitive impairment are associated with more widespread BOLD signal changes than absences with no or only mild cognitive impairment (Berman et al., 2010). However, in a case report of a girl with long-lasting GSWDs, paroxysms without concomitant cognitive impairment the same networks were activated as in absences with clinical manifestation (Moeller et al., 2010b).
As mentioned before, GSWDs are associated with an involvement of the thalamocortical network. However, is this network related only to GSWDs or do general alterations in this network exist in generalized epilepsy? Studying children with absence seizures, Bai et al. (2011) found increased interhemispheric connectivity in the orbitofrontal cortex, indicating increased synchronous activity between both hemispheres at rest. Increased connectivity was also found in the network of the basal ganglia when compared to healthy controls. This increased connectivity was even more pronounced during periods with GSWDs (Luo et al., 2012). However, decreased functional connectivity was described for the thalamus (Masterton et al., 2012; Wang et al., 2012), the DMN (Luo et al., 2011), and the attention network (Killory et al., 2011) in children with absence seizures. These findings might explain impaired interictal attention in these children. Decreased connectivity in the DMN areas was negatively correlated with epilepsy duration (Luo et al., 2011). Furthermore, Yang et al. (2012) showed that this decreased connectivity within the DMN, the cognitive control network, and the attention network was more pronounced during interictal GSWDs. However, a functional connectivity study in adult patients with different types of IGE did not show pathologic connectivity in areas that are known to interact during GSWDs (Moeller et al., 2011b). The studies on functional connectivity need further replication.
It is noteworthy that the same thalamocortical network is associated with primary and secondary GSWD paroxysms, with ictal, and interictal generalized discharges independent of age and antiepileptic medication. However, not all generalized discharges show involvement of the thalamocortical network. Photoparoxysmal response (PPR) is an electroencephalographic trait characterized by the occurrence of generalized epileptiform discharges in response to visual stimulation. An EEG-fMRI study in children with PPR showed involvement of the parietal and frontal cortex but not the thalamus in most of the subjects during generalized PPR (Moeller et al., 2009a). It seems likely that PPR is mainly a cortical phenomenon. However, in a patient, in whom photic stimulation evoked a generalized tonic–clonic seizure, an excessive increase in BOLD signal was detected in the visual cortex together with the involvement of the thalamus (Moeller et al., 2009c). These results were confirmed in a source analysis study in which dynamic imaging of coherent sources (DICS) was performed in the EEG studies recorded in previously mentioned EEG-fMRI studies in absences and PPR (Moeller et al., 2008b, 2009a). As revealed in the EEG-fMRI studies, DICS showed a thalamic source in all absence patients and in none of the PPR patients. Just like in the EEG-fMRI analysis, however, a thalamic source was found when PPR preceded a generalized tonic–clonic seizure. Concordant results between EEG-fMRI and DICS were also found for the DMN in the case of absences and for the occipital cortex, the parietal, and premotor regions in PPR. Valuable additional information could be derived from the analysis of the interaction between the sources based on renormalized partial directed coherence (RPDC). In absences, RPDC demonstrated an information flow from the prefrontal sources to the parietal sources on the cortical level. However, all cortical sources were similarly influenced by the thalamic source. In PPR, the flow of information between the cortical sources was dominated by the opposite posterior-anterior direction: from the occipital cortex, via the parietal cortex, to the frontal cortex (Moeller et al., 2012).
In summary, EEG-fMRI represents a valuable method for studying pathogenetic mechanisms of ictal and interictal GSWDs. EEG-fMRI was not only able to describe a specific fingerprint of activation associated with GSWDs but was also helpful in investigating the temporal dynamics of paroxysms and provides novel insights into epileptogenesis of IGE.
Epileptic encephalopathies (EEs) refer to a group of severe epileptic disorders characterized by recurrent seizures that result in cognitive and behavioral disturbances. In most cases, EEs are difficult to treat and are associated with an unfavorable outcome. However, very little is known about pathogenetic mechanisms or neuronal networks underlying EEs. One important feature of most EEs is that different etiologic factors, such as genetic background, lesions, malformations, and neurometabolic abnormalities, result in a similar clinical picture, such as similar type of seizures or similar interictal abnormalities. It might be possible that different etiologies at a certain developmental stage engage encephalopathy-specific pathogenetic pathways that constitute specific neuronal networks responsible for each particular EE. Based on this suggestion, different fMRI studies have been conducted to specify fingerprints of EEs on the network level, which could explain clinical similarity despite etiologic heterogeneity.
West syndrome is a prototypical severe epileptic encephalopathy of infancy consisting of spasms, psychomotor developmental delay, and a characteristic EEG pattern of hypsarrhythmia, which is characterized by multifocal spike and sharp wave activity and high voltage slow waves (Hrachovy & Frost, 2003). Early PET studies revealed a network underlying hypsarrhythmia, which included brain areas of focal cortical hypometabolism and subcortical hypermetabolism in the putamen and brainstem (Chugani et al., 1992; Chiron et al., 1993). It is unclear, however, whether this network is associated with epileptiform discharges or high-amplitude slow activity, since both pathologic phenomena may be epileptogenic in West syndrome. In an EEG-fMRI study, Siniatchkin et al. (2007a) investigated infants with West syndrome and demonstrated that epileptiform discharges cause positive BOLD signal changes in the cerebral cortex (especially in the occipital areas), whereas the high-amplitude slow-wave activity in hypsarrhythmia is commonly associated with BOLD signal changes in the brainstem, putamen, and thalamus (Fig. 4). There were no negative BOLD signal changes associated with the hypsarrhythmia, especially no deactivations in the DMN. Moreover, the activation pattern seemed to be specific for West syndrome, as it was not observed in infants with focal epilepsies. A subsequent source analysis, which was performed for high-amplitude slow activity in the same sample of subjects, revealed similar electrical sources in the occipital cortex, putamen, and brainstem as shown by fMRI, and even described the functional hierarchy between the sources: the activity in the brainstem seemed to influence other sources and may represent a key pathogenetic feature of West syndrome (Japaridze et al., 2012). The involvement of the brainstem can explain the clinical seizures that might result from intermittent interference of descending brainstem pathways controlling spinal reflex activity. Furthermore, ascending brainstem pathways that project widely to the cerebral cortex could explain characteristic EEG features (Frost & Hrachovy, 2005).
It seems likely that the activity in the brainstem represents the common pathogenetic pathway in West syndrome and Lennox-Gastaut syndrome (LGS). West syndrome often evolves into LGS, an epileptic encephalopathy characterized by different types of seizures (tonic, tonic–clonic, and atonic seizures, as well as atypical absences), typical EEG changes (slow spike-wave complexes ranging from 1 to 2.5/s, runs of rapid spikes, and poly spikes), and accompanying mental retardation (Arzimanoglou et al., 2009). An EEG-fMRI study in children with both symptomatic and cryptogenic cases of LGS revealed activation of the brainstem and thalamus associated with epileptiform discharges (Fig. 4, Siniatchkin et al., 2011). Because the brainstem exerts control over the gating function of the thalamus through its influence on the reticular thalamic nucleus, activity in the brainstem (reticular formation and raphe serotoninergic pathways) and the thalamus may lead to diffuse changes of cortical excitability, which predispose the neocortex to multifocal epileptic activity (Hrachovy & Frost, 2003). A recent study described increased functional connectivity between brain areas belonging to the DMN in children with multifocal epileptic activity, including young patients with the LGS (J. Moehring, B. Kroeher, A. Galka, G. von Ondarza, F. Moeller, S. Wolff, O. Granert, E. Steinmann, O. Jansen, R. Boor, U. Stephani, M. Siniatchkin, unpublished manuscript). This increased functional connectivity might indicate that the brain of patients with LGS is prone to increased synchrony, which predisposes these patients to multifocal epileptic activity. In addition, a study of Siniatchkin et al. (2011) demonstrated that especially the centromedian and anterior part of the thalamus is part of a network specific for LGS. A significant role of thalamocortical pathways in LGS had been demonstrated previously: centromedian thalamic nucleus stimulation in patients with refractory LGS was an effective treatment strategy (Velasco et al., 2006). The anterior thalamus has been frequently and successfully used as a target for deep-brain stimulation in patients with multifocal and secondary generalized epileptic activity (Samadani & Baltuch, 2007). These therapeutic options underscore the significance of the thalamus in the pathogenesis of LGS and multifocal IEDs.
The thalamus was also found to be part of an epileptic network in the EE with continuous spike and waves during slow sleep (CSWS; Siniatchkin et al., 2010) and might explain EEG features of multifocality and the tendency toward generalization in CSWS. CSWS is an age-related disorder characterized by the presence of interictal epileptiform discharges during at least >85% of sleep and cognitive deficits associated with this EEG pattern. Despite etiologic heterogeneity, patients with CSWS were characterized by activation of a similar neuronal network (the perisylvian region, insula, and cingulate gyrus; Fig. 4), as well as deactivation in the DMN in all patients (Siniatchkin et al., 2010). A comparison with the electrical source analysis results suggested that the activations corresponded to both initiation and propagation pathways of epileptiform discharges in CSWS. Deactivations in structures of the DMN (Fig. 4) are consistent with the concept that epileptiform activity interferes with normal brain function by inducing repetitive interruptions of neurophysiologic function (Gotman et al., 2005).
Dravet syndrome, or severe myoclonic epilepsy of infancy (SMEI), is a difficult-to-treat epileptic encephalopathy of early childhood that is caused by a mutation in the SCN1A gene in 80% of patients. In a recent EEG-fMRI study, 10 carriers of mutations in the SCN1A gene were investigated (Moehring et al., 2013). However, despite the common etiology, the study revealed different individual neuronal networks underlying the interictal epileptiform discharges. The only common feature of brain activation consisted of positive BOLD signal changes in the DMN areas found in seven patients. However, even these activations were inconsistent, and the group analysis did not yield any results. Even if the attempt to describe a common syndrome-specific network for SMEI patients was not successful, the results correspond with those of other neuroimaging studies, which have demonstrated heterogeneous results in SMEI patients (Ferrie et al., 1997).
EEG-fMRI in pediatric epilepsies is a promising tool for the investigation of neuronal networks involved in the generation of IEDs. The technique was successfully applied in awake, sleeping, or sedated children with focal and generalized epilepsies and in all age groups. Advanced methods of spike detection, motion correction, and the application of individual HRFs should improve the sensitivity of EEG-fMRI studies in the pediatric population. Similar results were obtained in small groups of subjects during wakefulness and sleep, in medicated and nonmedicated children, and in children with or without sedation. However, the effects of sleep, sedation, and antiepileptic medication need to be investigated more systematically in larger groups of patients, since their impact on the sensitivity of EEG-fMRI and the localization of BOLD responses remains unknown.
EEG-fMRI studies in the pediatric epilepsy population showed that EEG-fMRI allows the investigation of neuronal networks involved in different types of epileptiform activity. We showed, for instance, that EEG-fMRI is able to describe specific fingerprints of brain activation for a number of epileptic encephalopathies. The combination of EEG and fMRI helps to study brain function in a variety of epileptic diseases and provides insight into pathomechanisms of epileptic activity. Finally, the clinical value of EEG-fMRI has to be validated in larger studies by comparing EEG-fMRI data with other investigative techniques (intracranial EEG recordings, functional neuroimaging methods [PET, SPECT], EEG source analysis) and with the postsurgical outcome.
This work was supported by grants of the German Research Foundation (DFG) (grant SI 1419/2-1 and SFB 855 [and D3]) and a Gerok Grant from the Christian-Albrechts-University Kiel.
None of the authors has any conflict of interest to disclose. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.