Objective This prospective, bicentre, blinded, intention to treat study assessed the clinical added value of magnetic source imaging (MSI) in the presurgical evaluation of patients with refractory focal epilepsy (RFE).
Methods 70 consecutive patients with RFE (42 men; mean age 31.5 years, range 3–63) from two Belgian centres were prospectively included. All patients underwent conventional non-invasive presurgical evaluation (CNIPE) and a whole head magnetoencephalography recording (Elekta Neuromag). Equivalent current dipoles corresponding to interictal epileptiform discharges (IED) were fitted in the patients' spherical head model and coregistered on their MRI to produce MSI results. Results of CNIPE were first discussed blinded to the MSI results in respective multidisciplinary epilepsy surgery meetings to determine the presumed localisation of the epileptogenic zone and to set surgical or additional presurgical plans. MSI results were then discussed multidisciplinarily. MSI influence on the initial management plan was assessed.
Results Based on CNIPE, 21 patients had presumed extratemporal epilepsy, 38 had presumed temporal epilepsy and 11 had undetermined localisation epilepsy. MSI showed IED in 52 patients (74.5%) and changed the initial management in 15 patients (21%). MSI related changes were significantly more frequent in patients with presumed extratemporal or undetermined localisation epilepsy compared with patients with presumed temporal epilepsy (p≤0.001). These changes had a clear impact on clinical management in 13% of all patients.
Conclusion MSI is a clinically relevant, non-invasive neuroimaging technique for the presurgical evaluation of patients with refractory focal epilepsy and, particularly, in patients with presumed extratemporal and undetermined localisation epilepsy.
Statistics from Altmetric.com
Magnetic source imaging (MSI) is increasingly used in the non-invasive presurgical evaluation of patients with refractory focal epilepsy to better approach the presumed location of the epileptogenic zone (PLEZ). MSI is a combination of magnetoencephalography (MEG) and structural cerebral MRI that estimates the location of electrical sources at the origin of magnetic fields recorded by MEG.1 2 This technique allows the study of brain function with a temporal resolution at the level of the millisecond and a spatial resolution of a few millimetres.1 2 Such spatial resolution can be achieved using a high number of sensors (275–306 sensors in modern whole head MEG systems) and by the fact that magnetic fields, as opposed to electrical currents, suffer minimum attenuation and distortion from the different tissues they have to cross to reach the scalp surface. MSI is therefore recognised as having better spatial resolution than EEG.1 2
Several studies have shown that MSI is a sensitive technique for detection and localisation of epileptic discharges; its average detection rate is about 70% in large patient populations.3–5 MEG is highly sensitive to neural sources that are tangential to the skull and, in comparison with EEG, is almost blind to radial sources and less sensitive to deeper sources.1 2 The heightened sensitivity of MEG to tangential sources explains why it can detect epileptiform discharges not captured by EEG (and vice versa) leading to better hypotheses about the PLEZ when used in combination.6–10 Compared with conventional non-invasive presurgical evaluation (CNIPE), a presurgical evaluation including MSI has been shown to provide clinically pertinent and complementary information about the epileptic disorder in about 20–30% of patients with refractory focal epilepsy, leading to better patient selection, increased accuracy of intracranial EEG (ICEEG) and potentially better postsurgical outcome.11–16 In a prospective study assessing the clinical added value of MSI, Sutherling et al showed that MSI results led to a change in the initial ICEEG planning in nine of 69 patients (13%) and changed the surgical decision in another 14 patients (20%), leading to a clear benefit in six of 69 patients (9%).15 However, this study only included patients with five or more epileptic discharges captured on MSI and, therefore, did not estimate the actual MSI clinical added value in the population of epilepsy surgery candidates.15 Using a similar prospective design but focusing only on patients who were candidates for ICEEG, Knowlton et al showed that MSI led to additional electrode coverage in 18 of 77 (23%) ICEEG cases with a seizure onset ICEEG involving these additional electrodes in 39%.16 However, a prospective study assessing the overall clinical added value of MSI compared with CNIPE in large groups of patients is lacking.
Using a blinded and intention to treat design, we prospectively assessed the clinical added value of MSI in a group of 70 consecutive patients with refractory focal epilepsy followed in two Belgian reference centres for the treatment of refractory epilepsy.
Patients and methods
Between September 2007 and April 2010 (September 2007 to September 2009, ULB-Hôpital Erasme; January 2009 to April 2010, Ghent University Hospital), all patients undergoing a presurgical evaluation of refractory focal epilepsy and who were not formally excluded from surgery after multidisciplinary discussion of CNIPE results, were prospectively included in the study. Patients formally excluded from surgery based on CNIPE results (ie, multifocal epilepsy or severe epileptic encephalopathy) were not considered in this study to avoid any bias in the estimation of the MSI clinical added value.
Based on these inclusion and exclusion criteria, 70 consecutive patients (28 women and 42 men; mean age 31.5 years, range 3–63) were included, of whom 47 patients were followed-up at the ULB-Hôpital Erasme (Brussels) and 23 patients at Ghent University Hospital.
The study was approved by the institutional ethics committees. Patients or legal tutors gave informed consent.
All patients underwent CNIPE that comprised clinical and neurological evaluations, neuropsychological assessment, prolonged video-EEG monitoring aimed at capturing habitual seizures (seizures not captured in two patients), structural MRI (1.5 T or 3 T), positron emission tomography with [18F]-fluorodeoxyglucose (67 patients) and Wada test (26 patients). MSI was performed at the time of CNIPE in all patients.
In a first step, results of CNIPE were discussed blind to the MSI results in respective multidisciplinary epilepsy surgery meetings (separate meetings for ULB-Hôpital Erasme and Ghent University Hospital and no faculty overlap) to determine the PLEZ and to set surgical or additional presurgical plans (A, focal resective surgery; B, ICEEG; C, rejected except if new decisive information from MSI). When ICEEG was planned, the depth electrodes or subdural grid(s) position was anatomically determined during the meeting. Participants involved in MSI analyses (XDT, EC) did not participate in the initial discussion about PLEZ hypotheses or in the decision making process. Based on PLEZ, patients were classified as having presumed temporal (pTemporal), presumed extratemporal (pExtratemporal) or undetermined localisation (temporal vs extratemporal) epilepsy.
In a second step, MSI results were presented and discussed multidisciplinarily. The surgical or additional presurgical plans, before and after integrating MSI information, were compared during the respective multidisciplinary epilepsy surgery meetings. Changes in patient initial management plans were considered relevant when MSI results led to: (1) a change in the initial plan (A, B or C); (2) a change in the initial ICEEG planning; or (3) additional non-invasive neuroimaging investigations aiming at further assessment of PLEZ or functional risks. Among these additional neuroimaging investigations, MSI guided high resolution 3 T MRI using a surface coil with limited field of view was available for patients included at the Ghent University Hospital from August 2009.17 This investigation was systematically applied to patients with equivalent current dipoles (ECDs) clustering in a brain region structurally normal on MRI.
In a third step, patients were followed-up to determine the clinical relevance of MSI based changes in the management plan.
MEG data acquisition
MEG data were acquired using the whole head 306 channel Elekta Neuromag system (MaxShield; Elekta Neuromag Oy, Helsinki, Finland) installed at the ULB-Hôpital Erasme, the characteristics of which have been described elsewhere.18 19
For all patients, spontaneous magnetic brain activity (eyes closed rest, supine position) was recorded over 1 h (bandpass 0.1–330 Hz, sampling frequency 1 kHz). Head position inside the MEG helmet was continuously recorded by means of four head position indicator coils. The locations of the head position indicator coils with respect to anatomical fiducials were determined with an electromagnetic tracker (Fastrak, Polhemus, Colchester, Vermont, USA). No specific paediatric MEG helmet was used.
Three paediatric patients who were unable to stay still during the measurements were sedated using chloral hydrate solution 5% (1 ml/kg).
Magnetic source imaging
Continuous MEG data were preprocessed offline using the signal space separation (SSS, 60 patients) or spatiotemporal SSS methods (tSSS, 10 patients, correlation coefficient 0.8, segment length 4 s) to suppress the residual interference and correct for head movements.20–22 The data were then bandpass filtered to 0.1–40 Hz and visually inspected for interictal epileptiform discharges (IEDs).18 19 23 Sharp signals (duration <200 ms) exceeding 150% of the background signal variance, seen on several neighbouring channels and producing clear dipolar magnetic field patterns, were considered as potential epileptic events.18 19 23 Events related to physiological artefacts or rhythms were rejected.18 19 23
Source localisations of epileptic events were obtained by conventional dipole modelling tools (Elekta Neuromag Oy) using spherical conductor models determined from patients' individual MRIs. ECDs were fitted at the onset (or the first valid dipole before the peak) and at the peak of every IEDs (no averaging) using, when required, a selection of at least 40 channels, including MEG sensors involved in the IED dipolar field pattern to optimise ECD spatial accuracy and avoid any influence of irrelevant magnetic signals.18 19 23 When required, multidipole modelling was used to discriminate IED origin from its propagation pathways. Dipole fits were considered valid when the goodness of fit was >80% and the 95% confidence volume was ≤20 mm3.18 19 23 ECDs were then superimposed on the coregistered patient MRIs.
Differences in IED detection and MSI related changes between patients with pTemporal, pExtratemporal or undetermined localisation epilepsy were statistically assessed using one way ANOVA. Results were considered significant at p<0.05.
Results of this study are summarised in figure 1 and detailed in supplementary table S1 (available online only).
CNIPE results blinded to MSI data
Based on CNIPE, 38 patients (54%) were classified as having pTemporal epilepsy, 21 (30%) as having pExtratemporal epilepsy and 11 (16%) as having undetermined localisation epilepsy. The initial management plan was A in 24 patients (34%), B in 31 patients (44%) and C in 15 patients (22%).
MEG and MSI results
MEG data were not interpretable in three patients (4%) due to magnetic artefacts not removed by tSSS. MEG did not show any IED in 15 patients (21.5%), and showed IEDs in 52 patients (74.5%). Among the 67 patients with interpretable MEG, PLEZ (pTemporal, pExtratemporal or undetermined) significantly influenced IED detection (p=0.04). Indeed, IEDs were detected by MEG in 67% of patients with pTemporal epilepsy, 95% of patients with pExtratemporal epilepsy and 81% of patients with undetermined localisation. Post hoc analyses revealed a significant difference in MEG IED detection between patients with pTemporal and pExtratemporal epilepsy (pTemporal vs pExtratemporal, p=0.02; pTemporal vs undetermined localisation, p=0.46; pExtratemporal vs undetermined localisation, p=0.28). Interestingly, 1 h of MEG investigation detected IEDs in one of the five patients in whom the prolonged continuous interictal EEG did not show IEDs (patient No 20).
Influence of MSI results on patient management plan
MSI results did not change the management plan in 55 patients (79% of all patients). When abnormal, MSI corroborated the initial plan in 95% of these cases and helped to pinpoint a non-identified focal cortical dysplasia in one patient (patient No 19).
MSI changed the management plan in 15 patients (21% of all patients, 28% of patients with abnormal MEG). Clinical data of those patients are detailed in tables 1 and 2. MSI related changes consisted of a change from C to B in three patients (figure 2A; patient Nos 2, 20 and 55), a change from C to A in one patient (figure 2B, patient No 44), a change in the initial ICEEG planning in five patients (figure 2C, D, patient Nos 7, 32, 36, 56 and 67) and in the planning of additional neuroimaging investigations in six patients (figure 2E, F; ictal single photon emission computed tomography in patient No 8; language functional MRI in patient No 31; and MSI guided high resolution MRI using a surface coil in patient Nos 42, 46, 57 and 58). MSI changed the management in 3% of patients with pTemporal epilepsy (one of 38 patients), 43% of patients with pExtratemporal epilepsy (nine of 21 patients) and 45% of patients with undetermined localisation epilepsy (five of 11 patients). The PLEZ (pTemporal, pExtratemporal or undetermined localisation) therefore significantly influenced the frequency of MSI related changes (p<0.001). Post hoc analyses revealed significant differences in the frequency of MSI related changes between patients with pTemporal epilepsy and both pExtratemporal or undetermined localisation epilepsy (pTemporal vs pExtratemporal, p<0.001; pTemporal vs undetermined localisation, p=0.001; pExtratemporal vs undetermined localisation, p=0.9).
MSI related changes in patient management plan occurred with the same proportion at the ULB-Hôpital Erasme (10 of 47 patients, 21%) and Ghent University Hospital (five of 23 patients, 21%).
Clinical relevance of MSI related changes
At completion of the study, sufficient follow-up was obtained in 12 of 15 patients (80%) to determine the clinical relevance of MSI related changes. MSI related changes had a clear impact on initial management in nine of these 12 patients (75%, 60% of the patients with MSI related changes, 13% of all patients; patient Nos 7, 20, 32, 36, 44, 46, 55, 56 and 58). In patient No 7, the additional right mesiotemporal depth electrodes confirmed the existence of an independent seizure onset zone, contraindicating resective surgery. In patient No 20, ICEEG confirmed the PLEZ in the right opercular region identified by MSI, but the patient refused resective surgery due to potential functional risks. In patient No 32, coverage of the right premotor and primary sensorimotor cortex with a subdural grid instead of using depth electrodes targeting the right frontal focal cortical dysplasia confirmed MSI based suspicion of seizure onset within an eloquent cortical area; this observation led to the patient's rejection of resective surgery. In patient No 36, the additional temporal depth electrodes confirmed the involvement of this region leading to a broader resection than initially planned, and the patient is now seizure free. Patient No 44, who was initially rejected for focal resective surgery, underwent left anterior temporal lobectomy with amygdalo-hippocampectomy and presents a clear reduction in seizures (4–6 seizures per month preoperatively, absence or 1 seizure per month postoperatively). In patient Nos 46 and 58, MSI guided high resolution MRI identified a subtle brain lesion not detected on previous conventional 3 T MRI (figure 2F). In patient No 55, the right central location of the epileptogenic zone suggested by MSI was confirmed by ICEEG and the patient underwent successful surgery combining multiple subpial transections and focal resection (seizure free for 3 months, daily seizures before surgery). In patient No 56, depth electrodes implanted in the left frontal gyrus rectus focal cortical dysplasia identified based on MSI results confirmed the involvement of this lesion which was resected, leading to complete seizure cessation. Therefore, MSI based management modifications had definite medical consequences in seven patients as new findings in patient Nos 46 and 58 have not yet produced decisive effects yet.
In the two patients in whom the MSI guided surface coil MRI investigation failed to identify any brain lesion, MSI based changes were not considered clinically relevant (patient Nos 42, 57), as in patient No 67 in whom MSI related change did not set the seizure onset zone location, although ICEEG confirmed the existence of an additional irritative zone in the brain area identified by MSI.
Finally, in three patients (Nos 2, 8 and 31), insufficient follow-up precluded assessment of the relevance of MSI related changes.
In this study, MSI changed the management plan in 21% of consecutive patients with refractory focal epilepsy who were candidates for epilepsy surgery. MSI related changes were significantly more frequent in patients with pExtratemporal or undetermined localisation epilepsy compared with patients with pTemporal epilepsy. These changes were clinically relevant in 75% of patients for whom follow-up was obtained at completion of the study. A clear benefit of MSI related changes was therefore observed in 13% of all patients.
Detection of interictal epileptic discharges with MEG
In this study, MEG detected IEDs in 74.5% of patients, a proportion similar to that reported in MEG studies involving large groups of epileptic patients.7 9 11 24 This IED detection rate obtained after 1 h of MEG is lower than the 93% detection rate by prolonged video-EEG monitoring in this population. Relevant comparison of MEG and EEG IED detection should involve time locked MEG and EEG data recording; such an investigation was not performed in this study. Some patients showed few IEDs on the prolonged video-EEG monitoring, which could potentially explain why we did not capture any IED in some of them with 1 h of MEG. Interestingly, the IED detection rate of MEG was significantly lower in pTemporal than in pExtratemporal epilepsies. This discrepancy was probably related to the high percentage of patients with presumed mesial temporal epileptogenic zone and the potential difficulty in detecting mesial temporal sources due to their deepness or radial orientation relative to the MEG sensor array. Among the five patients with normal prolonged interictal EEG, 1 h of MEG detected IEDs in one patient, with a major impact on clinical management. Although time locked EEG and MEG recording was not obtained in this patient, this finding confirms that MEG may detect IEDs not captured by EEG, leading to more reliable hypotheses about PLEZ in some patients.6–10
MSI related changes in patient management plan
MSI changed the management plan in 21% of all patients and in 28% of patients with abnormal MEG. Even if the type of MSI related changes was different between the two participating centres, the proportion of changes remained identical.
MSI related changes in surgical management were significantly more frequent in patients with pExtratemporal or undetermined localisation epilepsy compared with patients with pTemporal epilepsy. This is probably related to the high proportion (71%) of mesial temporal PLEZ among patients with pTemporal epilepsy. In these patients, MSI supported the initial management plan by confirming the absence of an extratemporal irritative zone, which increased the level of confidence regarding the decision making process. This is considered clinically valuable (although difficult to quantify) given the risks associated with resective epilepsy surgery. The high degree of concordance between MSI findings and CNIPE in patients with pTemporal epilepsy suggests that this technique could potentially replace prolonged video-EEG monitoring in some patients with this condition.19 25 This possibility needs to be addressed in prospective multicentre studies.19 Some studies suggested that MSI could be valuable in identifying anterior versus non-anterior temporal epileptic sources, helping in surgical procedure optimisation and outcome prediction.26 27 Data obtained in patients with both anterior and non-anterior ECDs who became seizure free after anterior temporal lobectomy or disconnection question the clinical relevance of this sub-lobar ECD classification (patient Nos 22, 48, 65 and 66).
Importantly, 93% of MSI related changes in the initial management plan occurred in patients with pExtratemporal or undetermined localisation epilepsy. Three main interrelated factors might explain this observation.
MEG detects irritative zones not captured by conventional EEG leading to better knowledge of the patients' epileptic disorder and better hypotheses about PLEZ (illustration in figure 2A).6–10 This particular benefit of MEG in extratemporal epilepsies probably stems from its better signal to noise ratio and higher sensitivity to tangential sources, in comparison with EEG.28
By pointing to specific brain areas, MSI leads to subtle lesion detection, either by re-examination of conventional brain MRI or on novel targeted structural MRI investigations (see patient Nos 19, 46, 56 and 58).
MSI helps to determine the focal property of the irritative zone and the relationship with structural lesions or functional cortices (see patient Nos 31, 32, 36 and 67). This improves patient selection, ICEEG planning accuracy and surgical management, as shown by this study and others.11–16
MSI related changes were considered clinically relevant in 75% of patients for whom a follow-up was obtained at completion of the study. These changes helped either to accurately locate the seizure onset zone (seven patients) or to identify a brain lesion not observed on conventional MRI (two patients). This had a clear impact on patient management by reorienting five patients previously rejected for surgery towards surgical or ICEEG procedures, leading to seizure cessation in one patient and seizure reduction in another patient. In two patients, MSI contraindicated surgery due to high functional risk, and in two others it correctly identified the seizure onset zone leading to resective surgery and seizure cessation.
The main aim of this study was to assess the clinical added value of MSI in the context of a typical presurgical evaluation programme. We did not compare MSI results with time locked EEG recordings or source reconstruction of interictal and ictal EEG epileptiform discharges. Such direct MSI and electrical source imaging (ESI) comparison clearly needs to be investigated in future prospective multicentre studies to determine how MSI enhances the presurgical evaluation compared with ESI. In particular, such a comparison would further specify the respective contribution of MSI and ESI for the detection of subtle MRI lesions, assessment of the focal nature of the irritative zone and location of the epileptogenic zone.
As discussed by Stefan et al,5 studies investigating the clinical added value of a localising test in the context of epilepsy surgery are subjected to limitations.
Even if presurgical evaluations are rather standardised across epilepsy surgery centres, surgical management and procedures are considered to be highly variable across centres depending on experience and expertise, which complicates relevant comparisons or generalisations. Our bicentre design partially addressed this limitation and the similar percentage of MSI related changes between the two centres involved strengthens our findings.
As observed in this study, some patients do not complete the presurgical evaluation to the point of either ICEEG or surgery, which bias the estimation of the test clinical added value.
It is usually difficult to independently quantify the real clinical added value of a single test due to the involvement of multiple factors in the multidisciplinary decision process.
The absence of a reliable reference standard (ICEEG, resection cavity, seizure outcome) to assess the clinical relevance of the test under consideration remains a major issue.
Even if this study is subject to these limitations, its rigorous design highlights the fact that MSI provides information that reintegrates on a presurgical track patients initially rejected for surgery, improves the yield of ICEEG planning leading to either successful resective surgery or unexpected rejection for resective surgery in some patients, and guides the selection of patients who should or should not proceed to surgical procedures depending on the postoperative functional risks. The additional information provided by MSI is therefore clinically relevant in the context of the presurgical evaluation of refractory focal epilepsy.5
The authors would like to thank Martine Lemaire, Mohamed Tahere and Fabienne Maes who participated in the MEG data acquisition.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Download Supplementary Data (PDF) - Manuscript file of format pdf
Funding This work was supported by research grants from Elekta (Elekta Oy, Helsinki, Finland), the ‘Fonds de la Recherche Scientifique’ (FRS-FNRS, 3.4604.07, Belgium), the ‘Service Public Fédéral, Politique Scientifique’ (Belgium) and the ‘Instituut voor de Aanmoediging van Innovatie voor Wetenschap en Technologie in Vlaanderen (IWT-Vlaanderen)’ (Belgium).
Competing interests The Laboratoire de Cartographie Fonctionnelle du Cerveau benefit from a research grant from Elekta Neuromag. XDT is ‘Clinicien Chercheur Spécialiste’ at the ‘Fonds de la Recherche Scientifique’ (FRS-FNRS, Brussels, Belgium). XDT received consulting fees from Elekta for consulting activities in 2010 and 2011. EC benefited from financial support from Elekta to participate at the annual congress of the American Epilepsy Society 2009. MB benefits from a research grant of the ‘Fonds pour la Recherche Industrielle et Agricole’ (FRIA, Brussels, Belgium).
Patient consent All patients signed informed consent forms which were approved by the respective ethics committees at the time of the study, after having read the study information sheets also approved by the respective ethics committees. These information sheets clearly stated that the data of the study will be published in international scientific journals and presented at international scientific conferences, and as far as possible the data have been fully anonymised. In addition, regarding BMJ policy for patient confidentiality (http://group.bmj.com/products/journals/patient-consent-forms): (1) none of the clinical data presented in this manuscript allows patient identification. The only demographic information provided is age and enrolment period in two Belgian medical centres. These personal data are not sufficient to identify the included patients; (2) patient information has been sufficiently anonymised (see above); (3) all included patients are still alive; (4) patient data are grouped and there are no case reports, anecdotes or photographs of any patient; (5–7) this manuscript does not include any patient photographs. Images of patient brain MRI are completely anonymised.
Ethics approval The study was approved by the ULB-Hôpital Erasme and Ghent University Hospital ethics committees.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.