Chapter 16 – Predicting the Outcome of Surgical Interventions for Epilepsy Using Imaging Biomarkers




Chapter 16 Predicting the Outcome of Surgical Interventions for Epilepsy Using Imaging Biomarkers


Clarissa Lin Yasuda , Ana Carolina Coan , Marina K. Alvim and Fernando Cendes



16.1 Introduction


It is still unknown why outcome is suboptimal in approximately 30–50% of operated patients.1 Unfortunately, even a comprehensive presurgical investigation cannot yet distinguish candidates who will benefit from surgical intervention from those who will persist with seizures.2 Yet, the unpredictability of ultimate seizure control may thwart more broad indication of surgical treatment. Although several high-quality studies have investigated possible outcome predictors in temporal lobe epilepsy (TLE), including combinations of clinical, electrophysiological, and multimodal imaging findings, a conclusive and definite presurgical biomarker capable of stratifying the most probable surgical outcome is still unavailable.


Considering extratemporal epilepsies, and more specifically focal cortical dysplasias, the uncertainties regarding surgical prognosis are even greater, since for most patients qualitative MRI is unremarkable.3 The preoperative investigation is usually more complex and relies on the combination of multimodal neuroimaging and electrophysiological methods.4



16.2 Magnetic Resonance Imaging


The available studies suggest quantitative MRI alterations of multiple limbic and extralimbic networks rather than a circumscribed lesion in the hippocampus are related to surgical outcome.2 Studies are conveying evidence that a favorable outcome is more likely to occur when cerebral abnormalities are particularly confined to the targeted area to be resected or disconnected.4



16.2.1 MR Volumetry


As hippocampal sclerosis (HS) has been considered the main source of seizure activity in TLE, different studies have attempted to extract morphological characteristics that could be associated with surgical outcome. While some studies associated good surgical outcome with presence of predominantly unilateral hippocampal atrophy,5 in a study of 87 patients, volumes of both ipsilateral and contralateral mesial structures (hippocampus and entorhinal cortex) were not associated with surgical outcome; neither was the volume asymmetry.6


It is now clear that the presence of a unilateral atrophic hippocampus cannot reliably predict surgical outcome for TLE-HS patients, regardless of concordant clinical and electroencephalography (EEG) data.



16.2.2 Hippocampal Shape Analysis


The investigation of more complex morphological characteristics of hippocampus, shape analysis, consistently revealed that patients with poorer surgical outcomes also presented with atrophy of anterior/posterior lateral aspect of the contralateral hippocampus.7 In a study of surface-shape analysis of mesiotemporal structures, the accelerated progression of atrophy of contralateral entorhinal cortex predicted poorer surgical outcome in TLE patients.8


The application of an unsupervised pattern-learning algorithm on subregional surface data of mesiotemporal structures (hippocampus, amygdala, and entorhinal cortex) in a large cohort of TLE patients yielded four homogeneous classes, each with a unique profile of abnormalities (TLE I = marked bilateral atrophy [68% Engel I], TLE II = ipsilateral atrophy [89% Engel I], TLE III = mild bilateral atrophy [65% Engel I], TLE IV = hypertrophy [44% Engel I]).9 By submitting these data to a linear discriminant analysis algorithm (LDA), it was possible to accurately predict surgical outcome in 81% of all patients (class-blind model, 89% of seizure-free patients and 59% of patients with persistent seizures), and in 92±1% of patients of each class separately (class-based models).9 Surface-based LDA performed better than conventional volumetry of hippocampus (73% of correct outcome prediction), entorhinal cortex (72% of correct outcome prediction), amygdala (71% of correct outcome prediction), and their combination (71% of correct outcome prediction).



16.2.3 Extra-hippocampal Morphometry


Previous morphometric studies have associated unfavorable outcomes with extrahippocampal abnormalities.2 From nonspecific findings of extrahippocampal abnormalities to quantification of gray matter atrophy with voxel-based morphometry (VBM),1012 persistence of seizures has been repeatedly associated with a widespread, extrahippocampal pattern of alterations. In the analysis of 49 left TLE patients with VBM of preoperative T1-weighted images, support vector classification was applied to the extracted maps and accurately predicted surgical outcome (94% for men, 96% for women). While in the male cohort larger WM volumes were associated to a good surgical outcome, in the female cohort relatively larger WM volumes were associated to a suboptimal surgical outcome.11 A logistic regression model showed that presurgical GM volume reduction outside the temporal lobes associated with presence or absence of mesial pathology classified correctly 85.3% of TLE patients with good outcome and 70% of those with poor outcome.10


Both morphometric13 and visual14 analyses have been helpful to predict surgical outcome in patients with neocortical epilepsies associated with MRI abnormalities suggestive of FCD with up to 90% sensitivity and 67% specificity,13 with better prognosis being associated with complete excision of the MRI lesion.14



16.2.4 Large-scale Network Abnormalities


Cortical analyses with thickness measurements in TLE-HS patients revealed an association between poorer outcomes and atrophy in temporopolar/insular cortices, while surgical failure in patients without HS was related to atrophy in the posterior quadrant.15 The investigation of large-scale whole-brain networks with graph theory applied to cortical thickness correlated suboptimal postoperatory seizure outcome with increased disruption of network topological organization (characterized by increased path length).16 Interestingly, the restricted analysis of topological properties of mesiotemporal structures preoperatively (based on surface-shape information collected from hippocampus, entorhinal cortex, and amygdala) revealed that a more segregated ipsilateral hippocampus (characterized by increased path length) predicted long-term seizure control. While restricted topological alterations of mesiotemporal networks may be characterized by an “isolated hippocampus” (easier to be surgically “neutralized”), the presence of large-scale network abnormalities may reflect a more widespread pattern of structural alterations that facilitates seizure relapse after surgery.17



16.3 Diffusion Tensor Imaging


As diffusion images provide information regarding directionality and magnitude of water diffusion in each voxel, they have been used to assess white matter properties and organization (more specifically about the fiber tracts and connections);18 it is therefore not surprising that a number of studies applying different methods (either voxel-based or tractography-based) uncovers temporal and extratemporal abnormalities in epilepsy.19 Although the association between white matter alterations and seizure frequency has not been fully explored, there has been increased interest in investigating its potential as a biomarker of seizure control.


Regression models revealed that both larger hippocampal asymmetry indices and ipsilateral apparent diffusion coefficient (ADC) measurement predicted optimal surgical results.20 Another study showed suboptimal surgical outcome in patients with significant bilateral fractional anisotropy (FA) reduction in thalamotemporal paths, even after correction for extent of resection.6


Despite the limitations of diffusion imaging to fully depict microstructural properties in white matter in both healthy and diseased states, diffusion MRI is currently the only method capable to assess fiber trajectories in vivo. Given the ability to appraise white matter architecture, it is not surprising that diffusion MRI tractography has been used to study structural connectivity, and ultimately applied in the construction of structural brain connectomes.21


In an analysis of preoperative diffusion MRI from 20 TLE-HS patients, the poor outcome group presented with altered graph theory properties in the subnetwork within ipsilateral temporal lobe (higher integration with lower segregation).22 In addition, they also presented with increased connectivity between ipsilateral medial temporal and both lateral temporal/parietal lobes, as well as between contralateral temporal pole and parietal lobe. In a more recent study of TLE patients (with and without HS),23 subnetworks from structural brain connectome (based on preoperative diffusion MRI) linking ipsilateral temporal with extratemporal regions were used to estimate surgical outcome. Good seizure control was more likely in patients with fewer alterations within this subnetwork (ipsilateral hippocampus, amygdala, thalamus, superior frontal region, lateral temporal gyri, insula, orbitofrontal cortex, cingulate and lateral occipital gyrus). Predictive models revealed that neural architecture predicted seizure-free outcome with 90% specificity (83% accuracy), while the combination with clinical data provided 94% specificity (88% accuracy). In another study of DTI brain connectome in TLE, a two-step connectome-based prediction model was applied to evaluate the potential of machine learning algorithm to predict surgical outcome.24 In addition to correctly separate patients from controls with 80% accuracy, the model predicted surgical outcome with 70% of accuracy (similar to “expert-based” clinical decision).


Presurgical DTI brain connectome in left TLE has recently been used in computational modeling of epilepsy and surgical simulation (with removal of nodes from patient networks and comparison of likelihood of postoperative seizure control). Persistence of seizures after simulated surgery was associated with contralateral and frontal abnormalities; in addition, a more patient-specific surgical approach (rather than a predefined standard temporal lobectomy) would potentially yield better outcomes.25


These results provide evidence for a potential role of presurgical connectome (combined with clinical data) as a biomarker of surgical outcome.



16.4 Magnetic Resonance Spectroscopy


Proton magnetic resonance spectroscopy (MRS) assesses neuronal integrity by quantifying the peak of N-acetyl-aspartate (NAA), a marker of neuronal integrity, usually by comparing its concentrations with choline or creatine peaks. Unlike MRI, single photon emission tomography (SPECT), and positron emission tomography (PET) techniques, only a few relatively large voxels are sampled with proton MRS. The relatively poor signal-to-noise ratio of proton MRS and the relatively long time required to obtain spectra makes this technique of limited clinical use.26


Comparisons with the EEG localization and surgical results have demonstrated that the reduced signal intensity of NAA can lateralize and localize the epileptogenic focus in patients with focal epilepsies, in particular TLE. However, these changes are often bilateral and the relative concentration of NAA can normalize after successful surgery in patients with mesial TLE.26



16.5 Positron Emission Tomography


Preoperative investigation in about one-quarter of refractory patients remains difficult,27 in particular when MRI is normal and/or discordant with EEG and clinical information. In this scenario, functional imaging like PET has an important role.28 Given PET’s “unique capability of imaging brain metabolism,”29 it has been used to aid localization of the epileptogenic zone,30 based on the differences of regional brain metabolism between ictal and interictal phases.30


The first image method employed in TLE evaluation was 18F-fluoro-2-deoxyglucose (FDG) PET,31 which measures regional glucose metabolism. It is still the most commonly used PET tracer in epilepsy and typically shows a reduced uptake inter-ictally.32


FDG-PET hypometabolism agrees in 56–90% with seizure onset (determined by intracranial EEG), and has a moderate to high sensitivity in localizing the seizure focus (71–89%). In patients with either normal MRI or conflicting data (MRI discordant with clinical/EEG data) FDG-PET indicated the correct localization of the seizure focus in 63% of TLE and 38% of frontal lobe epilepsy patients.29


While the presence of unilateral temporal hypometabolism33 or asymmetric glucose consumption in both temporal and uncal regions34 predicted better surgical outcome in TLE patients, the presence of extratemporal and bilateral temporal cortical hypometabolism were correlated with poor outcome.33 Wong et al. investigated the occurrence of extratemporal glucose hypometabolism in TLE patients and noted that it was more common in frontal lobe and insula, and more widespread in patients with secondarily generalized seizures, although not necessarily contiguous with the temporal lobe hypometabolism.35 These areas were named as “remote hypometabolism” by the author, and its extension was related to poor outcome.35 One study with nonlesional TLE patients revealed a good surgical outcome associated with greater proportion of resection of FDG-PET hypometabolism volume.36


Although MRI seems to be the most important exam to influence the surgical decision in TLE, in the follow-up, only FDG-PET yielded significant predictive value for surgical outcome in the study of Struck el al.28 In this study with 34 operated patients, multivariate analysis with logistic regression showed that FDG-PET was an independent predictor of surgical outcome, but not MRI or EEG. MRI-negative PET-positive TLE patients presented with very good surgical outcome, comparable to those with unilateral HS on MRI.37


Given the difficulties in detecting epileptogenic focus in extratemporal epilepsy, a multimodal approach is generally necessary for investigation; therefore, prognostic factors usually derive from combinations of different methods. For the investigation of epilepsy in tuberous sclerosis complex (TSC), the combination of FDG-PET/MRI (coregistered) with higher values of ADC presented higher accuracy in detecting epileptogenic tubers and adjacent cortical dysplasia, compared to the information extracted exclusively from structural MRI (tuber size) and DTI (lower FA values). This multimodal approach (FDG-PET, MRI, ADC) also yielded a good surgical outcome (9 of 11 patients became seizure-free after tuber and adjoining cortex removal).38 In a large study of neocortical epilepsy, the correct lesion localization by FDG-PET was an independent predictor of good outcome with an odds ratio (OR) of 2.49 (minimum of 2 years free of seizures), along with focal lesion on MRI (OR = 2.54) and localized ictal onset on EEG (OR = 2.37).39


In patients with focal cortical dysplasias, FDG-PET had a sensitivity of 95% to detect lesions and improved the surgical prognosis of patients with negative MRI.40 Localized FDG-PET hypometabolism was a significant and independent predictor of good surgical result.41 FDG-PET was also associated with better surgical outcome in patients with cryptogenic neocortical epilepsy when concordant with interictal EEG findings.42


Another useful radiotracer in epilepsy investigation is [11C]flumazenil (FMZ). It binds to the benzodiazepine site of g-aminobutyric acid (GABA) receptors32 and reflects the functional integrity of GABAergic system.43 Previous studies showed superiority of [11C]FMZ compared to FDG-PET to detect epileptic foci, marked by a more localized reduction of [11C]FMZ uptake in the epileptic foci. A suboptimal surgical outcome in TLE-HS patients was associated with increased ligand binding of FMZ in periventricular white matter (suggesting the presence of heterotopic neurons undetected by visual inspection),43 which was later confirmed in an independent cohort.32


In extratemporal refractory epilepsy poorer surgical outcome correlated with higher number of cortical areas with decreased FMZ binding, which were not removed with surgery. Although resection may not be required for all these remote lesions, a careful investigation of such cortical areas (even with intracranial EEG) may eventually add information for an optimized surgical outcome.44


α-[11C]methyl-L-tryptophan (AMT) PET was originally developed to evaluate the brain synthesis of serotonin. AMT-PE has been useful to detect the epileptogenic cortex even in the absence findings on MRI or FDG-PET. Early studies in patients with TSC showed an increased uptake of AMT in epileptogenic tubers,45 probably resulting from an increased metabolism via the inflammatory kynurenine pathway, rather than abnormalities in serotonin synthesis, suggesting an immune activation during the epileptogenic process.45 The resection of tubers with increased AMT uptake yielded better prognosis. Favorable outcome was also observed in patients with focal cortical dysplasia and AMT uptake, which was more associated with Type IIB FCD.46



16.6 Single-Photon Emission Computed Tomography


Single-photon emission computed tomography (SPECT) is useful to detect the epileptogenic zone, by determining the cerebral perfusion during ictal and interictal phases.30 The tracers used are 99mTc-hexamethyl-propyleneamine oxime (99mTc-HMPAO) or 99mTc-ethylcysteinate dimer (99mTc-ECD), which typically shows areas of hyperperfusion during ictal acquisition, reflecting a hypometabolism in areas associated with seizure onset or propagation.47


The comparison between interictal and ictal SPECT images is traditionally performed by visual analysis, although the accuracy of this method in extratemporal epilepsy is less satisfactory.48 An improved analysis method to analyze is based on the subtraction of interictal SPECT from the ictal image (coregistered with MRI [SISCOM]), which has consistently demonstrated better accuracy to localize seizure onset,48 in both temporal (70–90%)49 and extratemporal epilepsy,47, 49 including focal cortical malformations.48, 50 SISCOM analysis has been a predictor of good surgical outcome for both temporal and extratemporal epilepsies with unremarkable MRI findings.41


Despite ictal SPECT localizing cortical malformations (53–72% of efficacy) (including focal cortical dysplasia),48, 50 it is still contraversial how much of the ictal hyperperfusion has to be resected for optimal outcome. While one SISCOM analysis (15 patients) considered it unnecessary to completely resect the hyperperfusion cluster,47 another study reported better surgical outcome after complete removal of the hyperperfusion zone.50 The contradictory results may derive from difference in sample size and mostly from methodological differences, given a better delineation of epileptogenic zone by SISCOM.50


Despite the validation of SISCOM, it does not account for normal variation of voxel intensities between sequential scans from a subject. To compensate this random variation between images, some groups have modified the analysis performing postprocessing techniques with SPM (www.fil.ion.ucl.ac.uk/spm/), either coregistered with MRI (STATISCOM—statistical ictal SPECT coregistered to MRI)51 or without (ISAS—ictal-interictal SPECT analyzed by SPM).52 In a study with TLE patients (with and without HS), STATISCOM was superior to SISCOM for seizure localization.51 In addition, a higher probability of good surgical outcome was associated with a correct determination of TLE subtype by STATISCOM, 81% of patients with correct localizing STATISCOM were seizure-free, compared to 53% of patients with incorrect or indeterminate STATISCOM.51 A study of refractory epilepsy with normal MRI (mixed temporal and extratemporal epilepsies) confirmed superiority of both SPM-based analysis (ISAS and STATISCOM) to correctly localize surgical site, compared to SISCOM.53



16.7 Functional MRI


Functional MRI (fMRI) is based on the ability of paramagnetic deoxyhemoglobin molecules to create the BOLD (blood-oxygenation-level-dependent) contrast related to increased oxygen consumption during neuronal activity.54 fMRI can be used for both task-related and task-free (resting-state) acquisitions. In epilepsy surgery, task-related fMRI (i.e., motor or visual paradigms) has been widely utilized in presurgical evaluation with the aim to localize eloquent cortex.27 Functional connectivity extracted from resting-state fMRI is a method that evaluates interactions among brain regions at a time the individual is not performing any specific task.55 This allows the observation of patterns of connectivity of large-scale brain networks, which may be compromised in various neurological disorders, including epilepsy.56


To date, the major problem of resting-state fMRI use in the clinical practice of patients with epilepsy is the difficulty to obtain patient-specific results. Promising results have shown that single-subject resting-state fMRI connectivity can be concordant with the presumed epileptogenic zone as determined by clinical, EEG or other imaging information.56 However, no studies have adequately compared surgical outcome with fMRI connectivity at the individual level. Recently, in a group-level analysis, one study57 used clusters of hemodynamic changes correlated with interictal EEG discharges to perform fMRI connectivity analysis in patients with different types of epilepsies before surgery. They associated seizure recurrence after surgery with a less lateralized functional connectivity, suggesting that a high laterality of connectivity of the epileptogenic zone could be a predictor of surgical outcome. Although encouraging, these results have no clinical impact up to this point.


On the other hand, the combined use of multimodal neuroimaging techniques, such as electroencephalography (EEG) and fMRI (EEG-fMRI), allows individual-level analysis.58 EEG-fMRI provides noninvasive simultaneous evaluation of neuronal activity and cerebral hemodynamics through the variation of the BOLD signal, detecting hemodynamic changes related to pathological ictal and interictal neuronal activity; therefore, it aids the determination of brain region responsible for the initiation and propagation of seizures.59, 60 Distinctly from resting-state fMRI, some studies published in the last years have tried to evaluate the value of EEG-fMRI in predicting surgical outcome in patients with focal refractory epilepsies. However, it is noteworthy that so far the number of studies and patients evaluated are still limited and currently EEG-fMRI is not widely used in the clinical practice. Additionally, systematic evaluation of the potential of EEG-fMRI to define the surgical outcome has been performed only through retrospective studies.



16.7.1 Functional MRI in TLE


In TLE patients, one recent study tried to compare fMRI connectivity of patients with or without seizure recurrence after surgery at both group and individual levels.61 TLE patients with recurring seizures after surgery had differences in the connectivity patterns of the right hippocampus with the ipsilateral thalamus compared with seizure-free patients. However, this difference could be either an increase or a decrease in the connectivity values and the small sample size did not allow further conclusions for individual patients.


In a recent study of TLE, Coan et al.62 retrospectively evaluated the value of EEG-fMRI to predict the long-term outcome (mean follow-up of 46 months), demonstrating that the presence of interictal-related BOLD changes in the area of the surgical resection was associated with good surgical outcome (seizure freedom, except for auras); it provided sensitivity and specificity of 81% and 79%, respectively, and positive and negative predictive values of 81% and 79%, respectively (Figure 16.1). In a multivariate analysis, EEG-fMRI results were associated with good surgical outcome independently of etiology, epilepsy duration or the duration of the follow-up.62





Figure 16.1. Example of interictal-related EEG-fMRI activations acquired during the presurgical evaluation of two distinct patients with pharmacoresistant epilepsies. In both cases, the presurgical BOLD maps are shown coregistered with the postoperative MRI and one example of the interictal discharge observed during the MRI acquisition is shown (A1 and B1; blue arrow). In case A, the area of maximum BOLD activation (red arrow) coincides with the surgical resection and the patient is seizure-free after a follow-up of three years. In case B, there is no BOLD activation inside the surgical resection and the area of maximum BOLD activation (red arrow) was not removed; this patient remained with frequent seizures after surgery.



16.7.2 Functional MRI in Extra-temporal Lobe Epilepsy


No studies comparing single-subject fMRI connectivity and surgical outcome in patients with extra-temporal lobe epilepsy have been published to date, but a few EEG-fMRI reports have combined both TLE and ETE showing that individuals with preoperative interictal EEG-fMRI results discordant with the surgical resection presented with poor seizure control after surgery.60 However, it appears that interictal EEG-fMRI has a low sensitivity to detect the epileptogenic zone.63


An et al.64 evaluated the role of EEG-fMRI to predict surgical outcome in 35 consecutive temporal and extratemporal epilepsy patients and found that the concurrence of both maximum BOLD change and surgical resected area, compared to the absence of any BOLD signal inside or in the vicinity of the surgical lacuna, could predict seizure-free outcome with high sensitivity (87.5%), specificity (76.9%), positive predictive value (70%), and negative predictive value of 90.9%.



16.8 Electric Source Imaging


EEG measures the neuronal activity on a sub-millisecond time scale. However, it lacks the ability to accurately localize the responsible brain source due to the so-called inverse problem: different source arrangements can generate the same distribution of electrical potentials. The development of different inverse solutions algorithms in the last decade has allowed us to implement source localization of the electrical potentials obtained by EEG (electric source imaging [ESI]).65 ESI can be used to localize the generators associated with both interictal or ictal epileptiform discharges, although the majority of studies are based on interictal activity. Although ESI has an interesting appeal to be used as a clinical tool, especially because of its low cost comparable with other modalities, its interpretation requires caution due to the variability of methodologies used in the different studies. The models used to generate the source maps are based on complex mathematical assumptions that vary between different studies.66 In addition, the accuracy of the source localization is significantly increased with 128 or more electrodes, in comparison with the standard 31 electrodes montage.65


Various studies have demonstrated that ESI can estimate the source of epileptiform potentials, with good accuracy to localize the epileptogenic zone defined either by noninvasive65 or invasive data,67 in both adults and children with focal epilepsies. The concordance of EEG source localization and intracranial EEG results has been demonstrated in both temporal and extra-temporal lobe epilepsy.68, 69 More recently, different studies have addressed the role of ESI in predicting the surgical outcome in patients with refractory epilepsies.


Few ESI series focused specifically on TLE patients. Recently, TLE patients underwent investigation with high resolution (256-channels) ESI.70 The results showed high sensitivity (91.4% for sublobar localization and 97.1% for lobar localization) and specificity (75%) for localization of the epileptogenic zone. Patients with sources localized within the brain area resected had better surgical outcome (Engel I or II) than those who did not.70 In ESI studies combining both TLE and ETE patients, while some failed to show significant difference in its accuracy to predict the surgical outcome between TLE and ETE,71 others showed higher localization power for TLE patients.72


A series including both temporal and extra-temporal lobe epilepsy patients evaluated for surgery showed that high resolution (128-channels) ESI detected a focal epileptogenic region in 32 patients, which was highly correlated with the defined epileptogenic zone (94% of cases); in addition, the identified source colocalized with the resected region in 79% of patients (19 of 30 operated patients).65


A large prospective study included 152 operated patients with focal epilepsies (minimum 1 year of follow-up) who underwent ESI during the presurgical evaluation.73 Comparing the concordance of the source localization with the resected area and the surgical outcome, the authors observed that high resolution (128–256 electrodes) ESI had sensitivity and specificity of 84% and 88%, respectively.73


More recently, Russo et al. also compared the use of three-dimensional low resolution ESI in the presurgical evaluation of 60 children with refractory epilepsies (mixed temporal and extratemporal) with other neuroimaging techniques.72 In addition to information for surgical planning as an adjunctive tool, ESI showed a sensitivity of 65.9% and a specificity of 36.8% to determine favorable outcome (Engels I and II) at 1-year follow-up and 60.6% and a specificity of 50% at 2-year follow-up.72


Lascano et al.74 performed a prospective study of 58 consecutive operated patients who had been submitted to four different modalities to localize the epileptogenic zone (MRI, high density ESI, PET, and ictal SPECT). Only MRI and ESI were predictors of favorable outcome, with increasing predictive values if both modalities presented with positive results.74



16.9 Magnetoencephalography


Magnetoencephalography (MEG) is a noninvasive technique that measures the magnetic field produced by the electric currents flow generated by the brain neuronal activity.75, 76 The superimposed MEG data on an MRI is called magnetic source imaging (MSI). Currently, MEG systems provides high-resolution recordings (100–300 channels) and it has been approved by the FDA for clinical use since 1997.


MEG has higher sensitivity for spike detection and localization than scalp EEG and it can add information to scalp video EEG in patients with nonlocalizing findings.77 Studies comparing MSI with other modalities have demonstrated its ability to localize the epileptogenic zone, as well as its relationship with the structural lesions, in both adults and children with epilepsy.75 MEG can also map the motor and sensory cortex adding information to presurgical evaluation. To date, most MEG studies in epilepsy are based on interictal findings, although some reports show that ictal MEG might provide superior localization information than interictal MEG in some cases.78


The comparison of MEG and ictal intracranial EEG demonstrates that MEG exams with well-localized sources are highly concordant with the seizure onset zone determined by the intracranial recordings.79 Indeed, seizure-free outcome is more commonly observed if a single cluster is identified by MSI and is concordant with the ictal onset zone determined by intracranial EEG. Differently, multiple clusters most often correlate with extensive ictal onset zone and poor seizure control after surgery.80 Focusing more often in well-localized MEG clusters, different studies have compared MEG/MSI results with surgical outcome in patients with refractory epilepsies.


According to the current data, MSI greatest value is to localize the epileptogenic zone in patients with lateral TLE, whereas for mesial TLE its accuracy is reduced.41 However, different studies have focused specifically on TLE patients, showing MEG can add information to the surgical plan of these patients.81 Anterior temporal localization of MEG spikes is associated with good surgical outcome after anterior temporal lobectomy, while spikes localized in regions other than the anterior temporal lobe or with lobar spike localization appear to have worse surgical results.82


Another study evaluated MEG results of 131 patients submitted to epilepsy surgery. Considering only the group of patients with good surgical outcome (Engel I or II), MEG correctly identified the epileptic lobe in 87.3% of TLE (patients) and in 100% of ETE.83


In a group of five patients with lesional frontal lobe epilepsy submitted to epilepsy surgery, Genow et al. showed that those with the majority of MEG spikes localized within the resected area presented good surgical outcome.84




References


1.de Tisi J, Bell GS, Peacock JL, et al. The long-term outcome of adult epilepsy surgery, patterns of seizure remission, and relapse: a cohort study. Lancet. 2011;378:1388–95. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

2.Bonilha L, Keller SS. Quantitative MRI in refractory temporal lobe epilepsy: relationship with surgical outcomes. Quant Imaging Med Surg. 2015;5:204–24.Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

3.Bernasconi A, Bernasconi N, Bernhardt BC, et al. Advances in MRI for “cryptogenic” epilepsies. Nat Rev Neurol. 2011;7:99108. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

4.Yasuda CL, Cendes F. Neuroimaging for the prediction of response to medical and surgical treatment in epilepsy. Expert Opin Med Diagn. 2012;6:295308. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

5.West S, Nolan SJ, Cotton J, et al. Surgery for epilepsy. Cochrane Database Syst Rev. 2015;7:CD010541.Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar

6.Keller SS, Richardson MP, Schoene-Bake JC, et al. Thalamotemporal alteration and postoperative seizures in temporal lobe epilepsy. Ann Neurol. 2015;77:760–74. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

7.Lin JJ, Salamon N, Dutton RA, et al. Three-dimensional preoperative maps of hippocampal atrophy predict surgical outcomes in temporal lobe epilepsy. Neurology. 2005;65:1094–7. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

8.Bernhardt BC, Kim H, Bernasconi N. Patterns of subregional mesiotemporal disease progression in temporal lobe epilepsy. Neurology. 2013;81:1840–7. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

9.Bernhardt BC, Hong SJ, Bernasconi A, et al. Magnetic resonance imaging pattern learning in temporal lobe epilepsy: classification and prognostics. Ann Neurol. 2015;77:436–46. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

10.Doucet GE, He X, Sperling M, et al. Frontal gray matter abnormalities predict seizure outcome in refractory temporal lobe epilepsy patients. NeuroImage Clin. 2015;9:458–66. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

11.Feis DL, Schoene-Bake JC, Elger C, et al. Prediction of post-surgical seizure outcome in left mesial temporal lobe epilepsy. NeuroImage Clin. 2013;2:903–11. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

12.Yasuda CL, Valise C, Saude AV, et al. Dynamic changes in white and gray matter volume are associated with outcome of surgical treatment in temporal lobe epilepsy. NeuroImage. 2010;49:71–9. CrossRef | Find at Chinese University of Hong Kong Findit@CUHK Library | Google Scholar | PubMed

Only gold members can continue reading. Log In or Register to continue

Jan 3, 2021 | Posted by in NEUROLOGY | Comments Off on Chapter 16 – Predicting the Outcome of Surgical Interventions for Epilepsy Using Imaging Biomarkers
Premium Wordpress Themes by UFO Themes