The 20 chapters in this book offer a comprehensive review of currently available knowledge regarding epilepsy surgery in MRI-negative patients, and they suggest that we have entered a novel era for treating these patients with the most challenging epilepsies. Three primary messages in this book deserve to be emphasized.

The first message is that re-evaluation of MRI data, either through optimal acquisition (more sensitive sequence, higher-field strength), post-processing tools, coregistration with other modalities (especially FDG-PET), and expert reassessment, enhances the detection of a large proportion of MRI-negative pathologies, such as focal cortical dysplasias (FCD) (Chapter 3). Up to 89% of small FCD overlooked by experts could be detected by using sophisticated post-processing methods of 3D-T1 MR images, coupled with neural networks automatic detection (Chapter 3).

The second message relates to the conceptual framework of presurgical evaluation. In patients with refractory focal epilepsy whose MRI discloses a well-defined epileptogenic lesion, presurgical evaluation concentrates on refining the localization of the epileptogenic zone (EZ) to be resected. In MRI-negative patients, presurgical evaluation also needs to provide insights into the underlying pathology for several reasons: (i) the presence and type of underlying pathology have a strong impact on the chance of postsurgical seizure freedom, with optimal surgical outcome in focal cortical dysplasia (FCD) type II and end-folium sclerosis, and poorest outcome in patients with histologically normal resected tissue; (ii) the extent of the EZ varies with the underlying pathology, being usually circumscribed to a sulcus in MRI-negative FCD type II, and often much more extensive in FCD type I (Chapter 11); and (iii) interpretation of presurgical data, such as the presence or absence of a clear-cut interictal hypometabolism on FDG-PET, will differ as a function of the suspected pathology. In other words, the “what” appears as important as the “where” when contemplating surgical treatment of MRI-negative patients.

The third message relates to the classic but disputable notion that postsurgical outcome is uniformly poorer in MRI-negative than in MRI-positive patients. Whereas the majority of series of temporal and extra temporal lobe epilepsies lead to this conclusion, some evidence suggests that this may not hold true if candidates are carefully selected for surgery. For instance, patients with MRI-negative but FDG-PET-positive temporal lobe epilepsy (TLE) have a similarly high rate of successful postsurgical outcome as those with MRI evidence of hippocampal sclerosis (Chapters 4 and 14). Similar findings were reported for MRI-occult FCD whereby the presence of concordant focal hypometabolism was associated with rates of seizure freedom that are comparable to those observed in MRI-detectable FCDs (i.e., about 90% class I Engel outcome) (Chapter 4). Altogether, state-of-the-art presurgical evaluation of MRI-negative epilepsy suggests that the surgery is worth undertaking, provided that there is an appropriate understanding of the issues at stake.

The most frequent pathological findings observed in operated MRI-negative patients are: (i) normal tissue in any lobe; (ii) FCD type IIA and type I in neocortical epilepsies; and (iii) either subtle hippocampal sclerosis or end-folium sclerosis in mesial TLE (Chapter 19). The exact proportion of each of these findings is unknown, and the proportion largely depends on the recruitment bias at each epilepsy surgery center. In the large European Epilepsy Brain Bank series, 8% of all resected specimen were histologically normal (Chapter 19). Such negative pathology does not exclude good surgical outcome following TLE surgery, but negative pathology makes excellent surgical outcome very unlikely following extra temporal surgery (Chapter 15). Whether this observation primarily reflects erroneous identification of the EZ, specific characteristics of histologically normal EZ that would prevent effective surgical treatment, or both, is unknown. Nonetheless, the observation militates for selecting patients with indirect evidence of MRI-occult FCD for extratemporal lobe surgery.

The FDG-PET technique coregistered on MRI (PET-MRI) appears to be particularly useful for suggesting the presence of FCD (Chapter 4). Not only does PET-MRI correctly identify the location of MRI-occult FCD in about 80% of cases, it shows very focal and severe hypometabolism with clear-cut borders, which is a metabolic pattern that is highly suggestive of FCD in the majority of patients. When FDG-PET is normal in MRI-negative TLE, PET investigation of 5-HT_1A receptors seems to be clinically useful for confirming the temporal origin of seizures, and for predicting favorable postsurgical outcome despite the lack of detectable FDG hypometabolism (Chapter 4). However, only a limited number of centers have access to PET tracers of the serotoninergic system.

Progress in the post-processing of ictal SPECT, and in particular the development of SPM-based methods such as STATISCOM, helps to delineate the EZ at a subregional rather than at a lobar level (Chapter 5). This gain in spatial resolution offers a substantial advantage for investigating MRI-negative patients, minimizing the risk of unnecessarily large placements of intracranial electrodes (Chapter 5).

Similarly, advances in the mapping of interictal epileptiform discharges using either high-resolution EEG with electric source imaging (ESI), MEG and magnetic source imaging, or EEG-fMRI, can more precisely identify the so-called irritative zone (Chapters 6, 7, and 8). Whereas we appreciate that the irritative zone does not always coincide with the EZ, their usual overlap make all the above methods useful. Furthermore, when investigating MRI-negative neocortical epilepsies with the hope of detecting an MRI-occult FCD, the presence of a well-delineated spike focus provides indirect arguments in favor of FCD, whereas the lack of such focus makes FCD unlikely. However, the absence of MEG-detected spikes would more strongly affirm the unlikelihood of deep-seated FCD with undetectable scalp EEG spikes, because MEG is usually very sensitive in detecting MEG spikes from deep-seated FCD. It also appears that the irritative zone better coincides with the EZ in type II FCD than in other FCD types. In any event, recent studies demonstrate that ESI, MSI, and EEG-fMRI abnormalities overlap with the EZ in 80% to 90% of MR-negative patients, and that the larger the resection of the spike focus delineated by these techniques, the greater the chance of postsurgical seizure freedom (Chapters 6, 7, and 8).

Whereas neuroimaging and electrophysiological investigations play a major role in the presurgical evaluation of MRI-negative patients, one should not underestimate the value of ictal phenomenology. In fact, advancing knowledge in electroclinical correlations validated by intracranial EEG recordings is helping the identification of patterns of ictal clinical sequence that are more reliable and distinct than in the past (Chapter 2).

The majority of MRI-negative patients will require intracranial EEG investigation prior to surgical treatment. One exception might be nondominant mesial TLE where direct surgery appears a reasonable option. Conversely, in dominant TLE, the possibility of sparing a morphologically normal hippocampus if the EZ proves to selectively involve the neocortex, the entorhinal cortex, the temporal pole, or the amygdala, justifies testing the hypothesis with the appropriate intracranial EEG recording (Chapter 14). Indeed, verbal memory loss proved to be significantly worse following anterior temporal lobectomy on the side dominant for language in patients with normal MRI than in those with hippocampal atrophy (Chapter 20). Invasive EEG investigation also appears to be mandatory for extratemporal lobe epilepsy (Chapters 15 and 16). The respective advantages and drawbacks of grids versus depth electrodes is still a matter of debate (Chapters 10 and 11). However, there has been a recent trend in some major epilepsy surgery centers in shifting from the former to the latter (Chapter 11). Reasons underlying this recent boost in stereoelectroencephalograpy (stereo EEG or SEEG) include: (i) more accessible technology, thanks to the development of robot-guided stereotaxy; (ii) evidence of a lower rate of serious complications than previously thought, with 1% to 3% of intracranial hematoma, rarely resulting in permanent deficit or death (≤ 0.6%); (iii) greater sensitivity than grids in detecting deeply located MRI-occult FCD, which is often located at the bottom of the sulcus; (iv) possibility of performing thermolesions of the suspected epileptogenic zone at the end of the recording sessions, the impact of which on seizures may help to refine the resective surgical strategy; and (v) high likelihood to result in a firm conclusion whether or not resective epilepsy surgery should be performed (Chapter 11).

Whatever the type of invasive investigation, the presence of high-frequency oscillations (HFOs), either ripples or fast ripples, appears helpful for delineating the EZ (Chapter 12). As for source imaging of interictal spikes, the extent of resection of HFO-generating tissue positively correlates with surgical outcome (Chapter 12). The presence of a permanent or subcontinuous focal spiking activity strongly suggests an underlying type II FCD.

Other specific diagnostic issues need to be considered for temporal and extratemporal lobe epilepsy. In patients with negative MRI and suspected TLE, one needs to consider the possibility of temporal-plus epilepsy (TPE) or extratemporal lobe epilepsy mimicking TLE (Chapter 14). Temporal-plus epilepsy refers to an EZ that primarily involves the temporal lobe but extends to the adjacent structures such as the orbitofrontal cortex, the insula, the frontal or parietal operculum, or the temporo-parieto-occipital junction (Chapter 14). In such patients, standard temporal lobectomy will usually fail to control seizures, whereas a more extensive resection, guided by SEEG, leads to a class I Engel outcome in two-thirds of patients (Chapter 14). Currently, invasive EEG remains the only reliable way to distinguish TPE from TLE (Chapter 11). Extratemporal lobe epilepsy mimicking TLE (usually originating in the extratemporal posterior cortex) can be suspected on various grounds, including the lack of a classic FDG-PET pattern for TLE (Chapters 4 and 14). There is however very little data available regarding MRI-negative posterior cortex extratemporal epilepsy (Chapter 16).

MRI-negative FLE patients have been reported in several publications, typically showing half the rate of postoperative seizure freedom as compared to lesional FLE (Chapter 15). As previously emphasized, this unfavorable outcome is likely a result of the combination of inappropriate selection of patients showing no evidence of an underlying FCD, and a suboptimal placement of intracranial electrodes in patients whose MRI-negative FCD will thus not be probed. Whereas FCD may affect any frontal sulcus, specific attention should be paid to the depth of the superior frontal sulcus, which is often overlooked on neuroimaging and not directly assessable with subdural grids. Efforts should now be developed to define criteria for selecting appropriate candidates for invasive EEG, based on the presence and/or congruence of noninvasive findings (video-EEG, FDG-PET, ictal SPECT, ESI, MEG, EEG-fMRI, etc.).