Simultaneous EEG-fMRI can assist with the presurgical evaluation of patients that suffer from a (potentially) surgically remediable syndrome. This is clinically important, as roughly 30 % of epilepsy patients have or develop pharmacoresistant disease (Kwan and Brodie 2000), i.e., seizures that cannot be controlled after adequate treatment with two antiseizure medications (as defined by the current consensus definition of the International League Against Epilepsy (Kwan et al. 2010). As mentioned in the introduction, epilepsy surgery can be very effective in this group. The first randomized controlled trial on epilepsy surgery has shown that for a carefully selected cohort of adult mesial temporal lobe epilepsy (MTLE) patients, the number needed to treat (NNT) to render one patient free of seizures that impair awareness at 1 year after surgery is two, and the NNT to render one patient free of any seizures is three (Wiebe et al. 2001). To use an illustrative example provided by Wiebe (2004), compare these results to carotid endarterectomy, the widely used standard for symptomatic carotid stenosis, with an NNT of 15 to avoid one disabling ischemic stroke or death in patients with severe stenosis over a follow-up of 2–6 years (Cina et al. 2000). No surgical complications occurred in the randomized controlled trial, whereas others have reported that surgical complications occur in 11 % of patients and that 3 % of patients sustain new, permanent neurological deficits (Engel et al. 2003). This is out-weighted by the fact that successful surgery can reduce the risk of premature death roughly twofold (Bell et al. 2010) and, according to some reports, even lead to low mortality rates that do not differ from the general population (Vickrey et al. 1995; Sperling et al. 1999). A recent study by De Tisi et al. followed 615 adult epilepsy surgery patients over a median follow-up of 8 years and found that a favorable long-term outcome was achieved in about 65 % of cases, where 51 % of patients showed immediate and sustained seizure freedom and an additional 14 % achieved seizure freedom after a complex process of remission and relapse (de Tisi et al. 2011). Some 25 % can be weaned of medication and can be considered as cured, but 40 % of operated epilepsy patients still require antiseizure drugs (ASD) to maintain seizure freedom (Schmidt et al. 2004; Tellez-Zenteno et al. 2005). However, large-scale retrospective studies in MTLE patients (n = 376, 18 years of follow-up) have shown that even if ASD are continued, roughly 19 % of patients in this category can safely switch from poly- to monotherapy or reduce preoperative monotherapy dosages (Wieser and Hane 2003, 2004). Plausibly owing to seizure freedom and decreased ASD side effects, quality of life improves significantly in operated patients, even if they suffer from postoperative memory decline (Langfitt et al. 2007b). As a consequence, utilization of health-care resources and health-care costs drop in the first 2 years after surgery (Langfitt et al. 2007a), and long-term comparisons of medical and surgical costs between operated and non-operated pharmacoresistant epilepsy patients indicate that epilepsy surgery becomes cost-effective in as little as 7–8 years (King et al. 1997; Langfitt 1997), although this estimate can be as long as 35 years, depending on the precise implementation of the analysis (Platt and Sperling 2002). Overall, these findings indicate that epilepsy surgery can be highly beneficial in terms of providing relief from individual suffering, prevent further disability, and reduce the health-care costs of society as a whole (Begley et al. 2000). Pharmacoresistant patients should therefore be identified and treated with surgery as early as possible (Kwan and Brodie 2000; Engel et al. 2012). There is still a delay of 18–23 years between development of pharmacoresistance and referral to presurgical evaluation, although recent data also suggest that the trend might be slowly (but modestly) reversing (Haneef et al. 2010).

For epilepsy surgery to be successful, the tissue to be resected and its spatial relationship to functionally relevant (eloquent) cortex have to be identified with precision. This goal requires a synthesis of data from multiple clinical and paraclinical sources of information, i.e., patient history, seizure semiology, neuropsychological examination, ictal and interictal scalp and/or intracranial EEG, and functional and structural neuroimaging (Rosenow and Luders 2001). Whereas standard fMRI with motor, sensory, or cognitive paradigms is used to map eloquent cortex, simultaneous EEG-fMRI is used as an additional tool to assist in mapping the epileptogenic zone (EZ), i.e., “the minimum amount of cortex that must be resected (inactivated or completely disconnected) to produce seizure freedom” (Luders et al. 2006). As mentioned in the introduction to this chapter, the EZ is thought to overlap with the seizure onset zone (SOZ) and the irritative zone (IZ) that generate seizures and IED, respectively. The IZ can thus serve as a proxy of the SOZ and the EZ for clinical purposes, although the different zones must not necessarily be concordant (Luders et al. 2006). Every neuroimaging technique used in epileptology has its strengths and weaknesses when trying to map and disentangle this complex topography. Scalp EEG, for instance, has a very high temporal resolution but is relative insensitive to deep sources of epileptiform activity within the EZ. PET and SPECT require invasive application of radioactive tracers. Structural MRI has a high spatial resolution and can indentify even subtle cortical epileptogenic lesions, but these must not overlap with either the IZ or SOZ. Simultaneous EEG-fMRI seems advantageous in this regard, as it can show the metabolic correlates of deep-lying IZs (via IED on scalp EEG) and depict the relationship between IED focus and epileptogenic lesion (via coregistration with high-resolution anatomical images). This might be especially important in those cases where mapping of the different zones is complicated or equivocal, e.g., in patients with the SOZ outside the temporal lobes, in those with multifocal lesions, and in those where no epileptogenic lesion can be identified by standard MRI protocols (Tellez-Zenteno et al. 2005; Jeha et al. 2007; Bien et al. 2009). A few studies have acquired simultaneous EEG-fMRI during the (fortuitous) occurrence of focal seizures (Sierra-Marcos et al. 2013), but safety concerns, movement artifacts, and the unpredictable nature of seizures limit the clinical applicability of this technique in our view.

There are now a number of clinical observational studies that indicate that local BOLD responses, as mapped with simultaneous EEG-fMRI, co-localize with the SOZ at the lobar level. Krakow et al. investigated 10patients with focal pharmacoresistant epilepsies (Krakow et al. 1999). BOLD correlates to IED could be found in 6, and in all of them the BOLD correlate was close to the SW maximum in the EEG. The largest study to date includes 63 consecutive patients (59) investigated with frequent IED (Salek-Haddadi et al. 2006). Concordant or approximately concordant BOLD correlates could be found in 23 patients (37 %); 25 had no IED and 11 no BOLD correlates, despite the presence of IED. This low yield might have had technical reasons since the study was performed at 1.5 T with spike triggering, and higher field strengths (Federico et al. 2005) as well as continuous imaging (Al-Asmi et al. 2003) lead to better results. Despite this shortcomings, this study revealed a few interesting results. For instance, there was a pattern of precuneal deactivation in seven patients, reminiscent of the patterns found in IGE (see above). Also, the authors found that BOLD activations showed in general a better concordance with the presumed SW focus than BOLD deactivations (the reason of which is still unknown). Finally, they found that most HRF had physiological waveforms, indicating that “standard” modeling approaches might be sufficient for clinical purposes. Kobayashi et al. investigated 35 patients with temporal lobe epilepsy and found concordant responses in 83 % (Kobayashi et al. 2006). Deactivations were again identified in roughly 50 % of cases. Interestingly, these authors found 16 patients with neocortical BOLD correlates of which 12 had concordant bilateral temporal lobe clusters, indicating that simultaneous EEG-fMRI does not only reveal EZ but can depict parts of the complete epileptogenic network. In a follow-up study, these authors found that this network extended not only to bilateral mesial temporal structures but also to interconnected areas such as the basal ganglia, inferior insula, and lateral temporal gyri (Kobayashi et al. 2009). Of note, contralateral temporal lobe BOLD activity peaked later in comparison to the ipsilateral activity. Thus, these results demonstrate that simultaneous EEG-fMRI, despite the low temporal resolution of the latter, might also show aspects of the temporal dynamics of SW-associated functional networks.

One important question as to the clinical utility of simultaneous EEG-fMRI is how closely it matches with two important gold-standard measures: results from intracranial EEG and postsurgical outcome (Zhang et al. 2012). This question has yet to be addressed prospectively, but there are several reports that provided interesting data on this subject. Seeck et al. provided an early case study of a patient in which simultaneous EEG-fMRI activity was concordant with scalp EEG source analysis and electrocorticography (ECoG) (Seeck et al. 1998). In the abovementioned case series by Krakow et al. in which six patients showed BOLD correlates of focal SW, 1 localization was also confirmed by ECoG (Krakow et al. 1999). Laufs et al. also found excellent concordance of simultaneous EEG-fMRI in one patient with a right frontocentral SOZ (Laufs et al. 2006). Interestingly, this patient did not show clear IED during the combined recording, and a focal abnormality of the EEG (localized delta activity) was used to derive the fMRI regressor. However, mixed results were reported in two of the largest studies to date by Bagshaw et al. (n = 4) and Benar et al. (n = 5) (Bagshaw et al. 2004; Benar et al. 2006). Both validated the findings of simultaneous EEG-fMRI against stereotaxic intracerebral EEG and found concordant BOLD activations to focal SW in 3/4 and 4/5, respectively, but also discordant BOLD correlates in one case per study.

To further clarify these issues, three recent studies have specifically asked whether simultaneous EEG-fMRI can add clinically useful information to the presurgical workup (Zijlmans et al. 2007; Moeller et al. 2009). Ziljmans et al. reviewed 29 patients with simultaneous recordings that had been rejected from surgery due to insufficient certainty in focus localization or multifocality. In their study, new prospects for presurgical reevaluation could be opened in 4, and two underwent intracranial studies that confirmed EEG-fMRI results. Moeller et al. studied another difficult group, patients with non-lesional frontal lobe epilepsy. They found BOLD correlates in 8/9. In two patients that were operated, subtle cortical abnormalities just adjacent to the EEG-fMRI correlates could be identified. A recent study by An et al. in 35 consecutive patients undergoing resective surgery investigated whether the peak of the BOLD response in the preoperative EEG-fMRI study matched the resected volume or not. Postsurgical seizure control was best in those patients were the EEG-related BOLD response was fully concordant with the resected area, whereas in patients with fully discordant BOLD responses, only 1 of 11 reached seizure freedom. This translated into a positive predictive value of the maximal EEG-fMRI response of 70 % and a negative predictive value of 90.9 %. This indicates that EEG-fMRI might be able to better identify those cases where surgery might not be successful.

In sum, the studies reviewed above testify to the complexity of results that can be obtained with simultaneous EEG-fMRI. The number of promising results predominates, and the fact that simultaneous recordings might be beneficial especially in difficult cases is encouraging. However, the fact that mapping epileptogenic tissue with simultaneous EEG-fMRI still mainly depends on the detection of IED is an important limitation. Also, adding to the complexity already afforded by the different overlapping zones (or rather networks and network components), the predictive value of IED to localize the SOZ is not always easy to determine and seems to depend on the neuroanatomical localization of the EZ and the pathophysiology of the underlying disease. For temporal lobe epilepsies, data indicate that IED on scalp EEG indeed predicts the SOZ correctly in ~90 % of cases (Blume et al. 1993, 2001), but for extratemporal sites, e.g., the frontal or parietal lobes, the predictive power of IED is lower, as only ~20 % of patients exhibits unilateral, clearly focal IED (Holmes et al. 2000). Thus, techniques that allow to map the EZ even without spikes are an important current focus of further research in EEG-fMRI (Rodionov et al. 2007; Grouiller et al. 2011). This could lead to an important new goal for simultaneous EEG-fMRI, i.e., guiding the placement of intracranial electrodes in patients with non-lesional or multifocal epilepsies (Zijlmans et al. 2007). Figure 4 shows one example that combines IED-related hemodynamic responses, post-operative imaging and intracranial EEG signals.

Epilepsy is a prototypical dynamic disease that includes a large-scale network and not only a circumscribed area of seizure generation. Normal brain function requires complex interactions between different, highly dynamic neural systems that rely on the integrity of structural and functional networks. Thus, a comprehensive evaluation of epilepsy necessitates not only the identification of the epiletogenic zone but the understanding of the functional embedding and interaction of this epiletogenic zone within an epileptic network. There is now emerging evidence that this view of epilepsy as a network disorder is more appropriate on anatomical, neurophysiological, and cognitive levels. Specific cortical and subcortical networks are increasingly recognized as fundamental elements that contribute to the generation and spread of focal onset seizures throughout the human brain. Network theory enables the interpretation of complex neural systems as an assembly of “nodes” and “edges,” representing functional and structural elements (brain areas) of a specific network and their structural and functional relationship. Functional network characteristics of the brain can be assessed by EEG and resting state BOLD fMRI either in combination or separately. While functional connectivity is considered as the physiologic process of “communication” between the brain areas within a network, structural network characteristics as measured by diffusion tensor or diffusion spectrum imaging “…can be considered as the “supporting hardwire” (van Diessen et al. 2013). Neurophysiologic studies have frequently reported increases in functional connectivity along large-scale networks in the brain and within the affected lobe that harbors the epileptic lesion (Bartolomei et al. 2006; Horstmann et al. 2010). This is in opposite to fMRI-based connectivity studies that reported widespread decreases in functional connectivity (Liao et al. 2010). These differences indicate differences in spatial and temporal resolution between neurophysiologic and fMRI studies. Simultaneous EEG and fMRI studies offer the potential to disentangle the relationship between neurophysiologic and BOLD-associated connectivity (van Dellen et al. 2009; van Wijk et al. 2010). A future goal of MR imaging would be also to measure changes of electromagnetic fields directly without the need of perfusion imaging or BOLD fMRI (Petridou et al. 2006). This technique, called neuronal current imaging, makes use of changes in ionic currents associated with synaptic and suprathreshold activity in the order of nanoamperes. While it has not reached its way into clinical practice, it would certainly lead to further insights into human brain functional organization (Bodurka and Bandettini 2002) and offer a new window for simultaneous recordings of EEG and MRI.