Address correspondence to Leonardo Bonilha, Division of Neurology, Department of Neurosciences, Medical University of South Carolina, 96 Jonathan Lucas St, 3rd floor CSB, Charleston, SC 29425, U.S.A. E-mail: email@example.com
Surgical resection of the hippocampus is the most successful treatment for medication-refractory medial temporal lobe epilepsy (MTLE) due to hippocampal sclerosis. Unfortunately, at least one of four operated patients continue to have disabling seizures after surgery, and there is no existing method to predict individual surgical outcome. Prior to surgery, patients who become seizure free appear identical to those who continue to have seizures after surgery. Interestingly, newly converging presurgical data from magnetic resonance imaging (MRI) and intracranial electroencephalography (EEG) suggest that the entorhinal and perirhinal cortices may play an important role in seizure generation. These areas are not consistently resected with surgery and it is possible that they continue to generate seizures after surgery in some patients. Therefore, subtypes of MTLE patients can be considered according to the degree of extrahippocampal damage and epileptogenicity of the medial temporal cortex. The identification of these subtypes has the potential to drastically improve surgical results via optimized presurgical planning. In this review, we discuss the current data that suggests neural network damage in MTLE, focusing on the medial temporal cortex. We explore how this evidence may be applied to presurgical planning and suggest approaches for future investigation.
Although >10 new antiepileptic drugs (AEDs) have reached the U.S. market in the past decade (Bialer, 2006), the most effective treatment for medial temporal lobe epilepsy (MTLE) continues to be the resection of the epileptogenic zone in the temporal lobe (Engel et al., 2003). Anterior temporal lobectomy is the standard of care for patients with MTLE who fail to achieve good surgical control with two first-line AEDs (Wiebe et al., 2001; Engel et al., 2003). With adequate presurgical planning and the correct definition of the seizure onset in the medial temporal lobe, patients with MTLE who undergo surgery may become free of disabling seizures in 65% of all cases, or have a significant improvement in up to 86% of all cases (Engel et al., 2003).
Unfortunately, around one of four patients with MTLE who undergo surgery fail to achieve seizure freedom despite state-of-the-art presurgical planning. It is disconcerting that there remain no ways to accurately predict who will or will not become seizure free. Before surgery, the patients who will not respond to surgery may appear identical to those who become completely cured of their epilepsy.
Many retrospective studies have attempted to answer whether surgical technique may influence outcome (Schramm, 2008). Although there is substantial variability in the results reported (possibly related to the fact that most studies evaluated a limited number of individuals), there are a few consistent points. Namely, there is evidence that selective amygdalohippocampectomy may be equivalent to anterior temporal lobectomy as far as seizure outcome (Arruda et al., 1996; Clusmann et al., 2002; Paglioli et al., 2006; Tanriverdi & Olivier, 2007), often with better neuropsychological performance after surgery (Helmstaedter et al., 2002; Clusmann et al., 2004; Tanriverdi & Olivier, 2007). Most studies have also found that a larger, nonspecific temporal lobe resection is not associated with better outcome (Jack et al., 1988; Bonilha et al., 2007b). However, the size of the resection of specific structures, which may be responsible for generating seizures independently, such as the hippocampus (Wyler et al., 1995; Bonilha et al., 2004b), the parahippocampal gyrus (Siegel et al., 1990), and the entorhinal cortex (Bonilha et al., 2007b), are directly related to seizure control. It appears that a larger temporal lobe resection is not always associated with better outcome, but a complete resection of the epileptogenic structures may be the key to surgical success. It remains unclear, however, how to predict before surgery which patients are more likely to benefit from a larger medial cortical resection.
It is notable that the majority of patients with MTLE who are currently operated become seizure free with conventional surgery (either temporal lobectomy or selective amygdalohippocampectomy), suggesting that for most patients the seizures start only within the hippocampus. However, there is a consistent rate of refractory patients regardless of the procedure chosen. These are the patients who are not identified as potentially refractory, for whom conventional surgery is not able to provide seizure freedom, suggesting that MTLE is not the same disease in all patients. Although for most patients the hippocampus is the primary epileptogenic area, for others the seizures may originate in adjacent structures, such as the entorhinal or perirhinal cortices. In these patients, complete resection of the hippocampus, but incomplete resection of the medial temporal cortex, may lead to suboptimal control of seizures. In support of this premise, it has been documented that 20% of patients with complete hippocampal resection confirmed by MRI continue to have seizures originating in the retained ipsilateral temporal structures (Wennberg et al., 2002).
MTLE may in fact comprise multiple subtypes, which appear similar clinically but have distinct prognoses. Therefore, the preoperative definition of the subtype of MTLE and the extent of the epileptogenic region is still an unmet clinical need. There are some related important unanswered questions: How can this information be translated into better surgical outcomes? Is there current evidence to suggest that we may be able to identify these subtypes of MTLE prior to surgery? Which presurgical diagnostic tools may be the most capable (i.e., sensitive and specific) of determining the extent of epileptogenicity?
In exploring these questions, we review evidence from distinct modalities, suggesting that MTLE is a network disease with a variable degree of extrahippocampal involvement, incorporating an illustrative case presentation. We then focus on the role of the medial temporal cortices as epileptogenic entities, and offer strategies for determining whether MTLE has distinct subtypes and how they may be applied to the presurgical evaluation.
MTLE as a Network Disease
There is accumulating evidence that MTLE is a disease that affects other brain structures in addition to the hippocampus. Data from structural MRI and scalp and intracranial EEG suggest that structural and functional compromise is not restricted to the hippocampus, but affects a network of structures within and outside the temporal lobes.
Over the last two decades, structural MR images (particularly high-resolution T1-weighted images) have been submitted to increasingly better techniques to quantify abnormalities that are not easily detected by the human eye. “Postprocessing” volumetric MRI techniques demonstrated that regional volume decrease in MTLE is usually linearly correlated with regional cell loss (Lencz et al., 1992).
There are two widely accepted forms of volumetric assessment of MRI images: manual morphometry and automated whole brain analyses. The first method involves the manual tracing of brain structures and computation of the volume by summing the areas of the two-dimensional traced regions (Jack et al., 1988; Bonilha et al., 2004a). This technique is highly effective; a large number of pioneering studies employed manual morphometry and provided significant insight into the pathophysiology of MTLE. For example, through the use manual morphometry, researchers have found that MTLE is associated with a quantifiable volume loss in the hippocampus (Jack et al., 1988), and the presence of unilateral atrophy is associated with a better surgical outcome (Arruda et al., 1996). Early studies with manual morphometry have also demonstrated that extrahippocampal structures are significantly atrophied in patients with MTLE (Bonilha et al., 2003). However, this technique is time-consuming and prone to rater bias. In addition, it does not enable an assumption-free analysis of the whole brain.
Postoperative imaging studies have suggested that the extent of the resection of the parahippocampal area, which usually comprises the entorhinal and perirhinal cortex in its most anterior portion (Insausti et al., 1998), leads to better outcomes when associated with complete hippocampal resection (Siegel et al., 1990; Bonilha et al., 2007b). However, it must be emphasized that these studies did not take presurgical imaging features into account. To date, there has never been a controlled prospective study evaluating whether patients with a higher degree of network damage are more likely to benefit from a broader resection.
Electrophysiologic data suggest that the entorhinal and the perirhinal cortices may be responsible for generating and maintaining seizures in some patients with MTLE (Spencer, 2002; Wennberg et al., 2002). In a recent review examining the pattern of EEG recordings from depth electrodes (Kahane & Bartolomei, 2010), the authors suggested that the large amount of evidence that seizures originate within the extrahippocampal temporal lobe structures contradicts the hypothesis that hippocampal sclerosis is the only determinant of epileptogenicity.
Distinct epileptogenic networks are involved in the generation of seizures in MTLE (Bartolomei et al., 2010). Within the temporal lobe, there may different types of seizure patterns, such as the mesial, mesiolateral, lateral, temporopolar, and temporal “plus” subtypes (Bartolomei et al., 2010; Kahane & Bartolomei, 2010). Experimental studies suggest that medial temporal fast (“ripple”) oscillations (140–220 Hz) and gamma oscillator (90–140 Hz) are quantitatively distinct patterns associated with hippocampal sharp waves. However, hippocampal sharp waves depend on CA3 region bursts, which are modulated by the entorhinal cortex, suggesting the existence of hippocampal “subnetworks” that are modulated by distinct excitatory input (Sullivan et al., 2011). Ripples and fast ripples oscillations have been demonstrated to exist not only in rodents, but also in humans (Le Van Quyen et al., 2008). Furthermore, data from microelectrodes suggest that cellular networks underlying fast ripple generation are more localized to the entorhinal cortex in patients with MTLE (Bragin et al., 2002). These distinct network patterns may appear identical from a seizure semiology perspective, but the functional dynamics and anatomic involvement are clearly distinct.
After standard presurgical evaluation, a right-handed 34-year-old underwent a left anterior temporal lobectomy for medication refractory MTLE at the age of 26 years. At that time, her MRI showed evidence of left hippocampal volume loss, and her ictal scalp EEG demonstrated left medial temporal onset of seizures. Unfortunately, her seizure frequency was unaffected by surgery (Engel class IV). Eight years after the original surgery the patient underwent a new comprehensive presurgical planning, this time including intracranial contact and depth electrodes. Her new MRI showed retained hippocampal structures (Fig. 1). With this finding, the immediate assumption might be that this was the source of onset of continued seizures. Nonetheless, during intracranial EEG the recordings clearly showed seizures originating in the entorhinal and perirhinal areas, rather than the hippocampus (Fig. 1).
This illustrative patient may not be an isolated case, as similar data has been described from other centers. For instance, using intracranial data, Wennberg et al. (2002) described that the parahippocampal cortex can originate independently and maintain seizures in patients with MTLE. Furthermore, in her classical paper on seizure networks, Dr. Susan Spencer suggested that the medial temporal cortices may play an important role in seizure generation and maintenance (Spencer, 2002).
The Medial Temporal Cortex as a Key Network Component
Converging evidence indicates that for some patients with MTLE, the extrahippocampal medial temporal cortex may play an important epileptogenic role. This is underscored mostly by: (1) the absence of seizure control after complete hippocampal removal; (2) data from MRI studies suggesting that this region is structurally abnormal in patients with refractory MTLE; (3) the demonstration by intracranial recordings that seizures originate in the entorhinal and perirhinal cortices in some patients with MTLE.
It is important to note that when patients who achieved seizure freedom are compared to patients with continued seizures after surgery, the location of the surgical resection is subtly but statistically different (Bonilha et al., 2007b). Patients who are free of seizures tend to have a more complete hippocampal and broader entorhinal–perirhinal resection, compared with patients who did not become free of seizures. Notably, the extents of resection of the hippocampus and the entorhinal cortex are independently associated with seizure freedom, and the overall size of surgical resection is not related to outcome (Bonilha et al., 2007b). Although the entorhinal and the perirhinal cortices may not be responsible for epileptogenesis in all MTLE patients, their probable role for a significant percentage of patients raises important yet unanswered questions. Should the entorhinal and perirhinal cortices be included in all temporal resections? Is there a trade-off to be expected between the extent of resection, seizure control, and residual postsurgical neuropsychological impairments?
Clearly, there are other possible explanations for suboptimal surgical results in MTLE, in addition to the involvement of the entorhinal and perirhinal cortex in epileptogenesis (Thom et al., 2010). Cumulative evidence suggests that other factors may influence outcome such as contralateral hippocampal involvement (Jack et al., 1992; Van Paesschen et al., 1995; Li et al., 2000; Lin et al., 2005), and the degree of extratemporal involvement (Yasuda, et al., 2010). The challenge is to correctly identify the patients for whom the parahippocampal structures play a significant role in seizure generation and maintenance, thereby optimizing treatment planning.
New Techniques Offer New Opportunities
New imaging and electrophysiologic techniques may help elucidate the extent of medial temporal lobe pathology and its relevance on epileptogenesis. In the near future, these data may guide a more individualized approach to surgical planning. In addition to data that can be obtained from intracranial EEG from selected patients, individualized morphometric medial temporal lobe MRI analysis may also prove to be very beneficial.
Quantification of hippocampal volume is more sensitive than the visual inspection of MRI to detect features associated with hippocampal sclerosis (Cendes et al., 1993), and volumetric atrophy is linearly related to cell loss in the hippocampus (Lencz et al., 1992). However, studies employing morphometric medial temporal lobe MRI have historically relied upon the manual delineation of the medial temporal lobe structures (Jack et al., 1990; Cendes et al., 1993), a time consuming, rater-dependent method. Furthermore, the reliability of the results requires a large database of normative, demographically matched control data. Despite these limitations, manual morphometry has been accepted as an additional diagnostic tool in some centers.
More recent MRI morphometric techniques such as voxel-based morphometry (VBM) or cortical thickness employ a computerized algorithm to quantify density, volume, and thickness of gray matter (Ashburner & Friston, 2000, 2001). The main advantages of automatic quantification of volume are the possibility to assess the whole brain quickly, the lack of anatomic limitations, and the absence of rater bias by manual measurement. These techniques, however, have become robust and dependable only with more recent improvements in the quality of the high resolution T1 sequences yielded by high-resolution high-field MR scanners.
Studies employing VBM to study the pattern of gray matter in patients with MTLE have consistently investigated the average atrophy in patients with MTLE (Keller & Roberts, 2008). It is possible to envision that VBM or other modalities of medial temporal lobe volume quantification may be used in the near future to assess the degree of entorhinal and perirhinal cortex atrophy in individual patients with MTLE. These techniques may be enhanced by the individualized analysis of the distribution of gray matter atrophy using standardized voxel-based maps of gray matter in single patients, which can be displayed as deviations from the average in matched controls (Bonilha et al., 2009). Because the analysis is unbiased and not limited to specific anatomic regions, extratemporal regions such as the insula and the cingulate cortex could also be objectively assessed. This information, combined with the degree of hippocampal atrophy, may lead to detection of patterns of atrophy that may guide the expectation of surgical outcome. For instance, patients with MTLE with a higher ratio of medial temporal lobe cortex atrophy to hippocampal atrophy may be hypothesized to have a higher chance for suboptimal outcome.
The use of automated morphometry for individualized structural abnormalities may be a promising tool in the presurgical assessment of MTLE. Given that automated methods rely heavily on image preprocessing and have been originally devised mostly for the evaluation of group effects, the methodology for individualized assessments is still under development. For example, in a recent study by our group, we observed that the likelihood of confirmation of clinically relevant hippocampal atrophy is more accurately determined through the analysis of all voxels comprising the hippocampus, but focusing on the deviation from the control mean of the voxels located in the bottom quartile (Bonilha et al., 2009). Further work is needed to refine the practical implications of this method, and automated techniques should not substitute the comprehensive assessment from experienced clinicians. Nonetheless, the use of automated quantification methods may prove to be a powerful decision support tool in the diagnosis of MTLE and the extent of structural pathology.
Similarly, the presence of more frequent interictal or ictal spikes detected may indicate a poorer outcome. The ratio of medial temporal spike frequency/hippocampal spike frequency may be used to assess disease severity. The results from a recent study employing scalp EEG suggest that the frequency of interictal temporal spikes are inversely correlated with outcome (Krendl et al., 2008). This suggests a higher “epileptogenicity” of the medial temporal lobe and may be an indirect evidence of the degree of damage of the hippocampus and entorhinal and perirhinal cortices. The increased frequency of spikes that are apparent on scalp EEG may indicate a more broadly affected (less-localized) epileptogenic region.
It remains to be determined whether the resection of the entorhinal cortex and the perirhinal cortex may lead to significant postoperative neuropsychological impairments. Recent studies combining MRI morphometric analyses and neuropsychological profile have shown that the perirhinal cortex is independently associated with verbal and general memory performance, at least in patients with left hippocampal sclerosis (Alessio et al., 2006; Bonilha et al., 2007a; Krendl et al., 2008). In theory, the resection of the medial temporal neocortex should perhaps be discussed with patients with presurgical evidence of entorhinal and perirhinal cortical damage. These patients may exhibit presurgical memory impairment, but it is as yet unknown if the resection of the medial temporal lobe may aggravate these deficits and perhaps lead to prohibitive levels of memory impairment. Future studies should carefully address the degree of change of pre- to postsurgical memory performance, in order to evaluate whether there is a neuropsychological cost of expanding the temporal lobe resection, and delineate the risk–benefit ratio from this approach.
In this article we suggest that novel MRI techniques may be used to reevaluate candidates for MTLE surgery. It has been well established that some surgically refractory MTLE patients continue to have seizures because the seizures originate in the medial temporal lobe cortex after surgery. Likewise, it has been demonstrated that resection of the medial temporal lobe cortex increases the chance of good surgical outcome. The use of preoperative high-resolution morphometric MRI may help discriminate the patients with a higher degree of pathology in the entorhinal and perirhinal cortices. The degree of pathology of the parahippocampal regions may indicate a greater likelihood of epileptogenesis from this area and it is possible that these patients could benefit from a more ample resection. Clearly, the involvement of the parahippocampal region is not the only theory to explain suboptimal surgical results, which may also be related to contralateral pathology or extratemporal neuronal loss. Nonetheless, parahippocampal epileptogenesis may play a significant role in seizure generation for some patients, and the current challenge lies in establishing new methods to determine the extent of parahippocampal or extratemporal pathology presurgically. Finally, careful neuropsychological studies should be planned to assess whether tailoring the resection to target damaged temporal lobe areas may have a positive or negative impact on memory performance after surgery.
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.