[Skip to Content]
[Skip to Content Landing]
Figure 1.  Localization of Seizure Onset Using Single Equivalent Current Dipole (sECD) Analysis
Localization of Seizure Onset Using Single Equivalent Current Dipole (sECD) Analysis

A, Detection and analysis of a seizure on a routine magnetoencephalography. The arrowhead indicates the seizure onset in the left parietal subset of contacts. The background is replaced by evolving admixed frequencies. B, Selection of a subset of magnetoencephalography sensors involved at seizure onset to increase the signal-to-noise ratio and efficient sECD fitting. The red circle is a schematic approximation of 30 to 60 sensors identified by a certified magnetoencephalographer and averaged for subsequent dipole fitting. The averaging is performed to maximize the contribution of the waveform of interest in dipole estimation and to eliminate/minimize other signals and weak fluxes contribution. C, Determining the fitting parameters of sECD of a spike at the onset of seizure. Top, The analysis was performed on the average magnetoencephalography signal segment (red) at a specific time point (vertical line). Bottom, The blue and red lines indicate outgoing and incoming magnetic fluxes, respectively. The green arrow indicates the location and direction of equivalent current dipoles. D, Coregistration of the results of sECD analysis after passing fitting parameters with T1-weighted magnetic resonance imaging (MRI). The yellow dot and line represent the location and the orientation of the dipole, respectively.

Figure 2.  L2 Minimum-Norm Estimate of a Narrow Band (MNE-fc)
L2 Minimum-Norm Estimate of a Narrow Band (MNE-fc)

A, A 10-second view of seizure onset in the right parietooccipital sensors. The low-frequency and high-frequency filters are set at 1 Hz and 40 Hz, respectively. The vertical line indicates the emergence of evolving paroxysmal fast activity at seizure onset. B, Morlet wavelet time-frequency decomposition of data. The red square indicates seizure onset. C, Demonstration of stability of current densities at the sensor level around the dominant frequency at onset (left) and in time with evolution (right). D, Concordant localization of seizure onset using single equivalent current dipole (sECD) analysis and L2 MNE-fc of seizure shown in A. E, The patient underwent 2 prior resections with early recurrences. The resection contained L2 MNE-fc activation but not the area pinpointed by sECD. The patient remains seizure free.24 The gray mesh indicates sites of previous resections. F, Modeling parameters of sECD were not met. Resection of the L2 MNE-fc area at the posterior edge of a previous resection led to seizure freedom. Color scales are normalized to the maximum current density per map in all panels.

Table.  Summary of Patient Demographic and Clinic Characteristics at the Time of Magnetoencephalography Recording
Summary of Patient Demographic and Clinic Characteristics at the Time of Magnetoencephalography Recording
1.
Knowlton  RC, Elgavish  R, Howell  J,  et al.  Magnetic source imaging versus intracranial electroencephalogram in epilepsy surgery: a prospective study.  Ann Neurol. 2006;59(5):835-842.PubMedGoogle ScholarCrossref
2.
Agirre-Arrizubieta  Z, Thai  NJ, Valentín  A,  et al.  The value of magnetoencephalography to guide electrode implantation in epilepsy.  Brain Topogr. 2014;27(1):197-207.PubMedGoogle ScholarCrossref
3.
Sutherling  WW, Mamelak  AN, Thyerlei  D,  et al.  Influence of magnetic source imaging for planning intracranial EEG in epilepsy.  Neurology. 2008;71(13):990-996.PubMedGoogle ScholarCrossref
4.
Knowlton  RC, Laxer  KD, Aminoff  MJ, Roberts  TP, Wong  ST, Rowley  HA.  Magnetoencephalography in partial epilepsy: clinical yield and localization accuracy.  Ann Neurol. 1997;42(4):622-631.PubMedGoogle ScholarCrossref
5.
Zimmermann  R, Scharein  E.  MEG and EEG show different sensitivity to myogenic artifacts.  Neurol Clin Neurophysiol. 2004;2004:78.PubMedGoogle Scholar
6.
Claus  S, Velis  D, Lopes da Silva  FH, Viergever  MA, Kalitzin  S.  High frequency spectral components after secobarbital: the contribution of muscular origin–a study with MEG/EEG.  Epilepsy Res. 2012;100(1-2):132-141.PubMedGoogle ScholarCrossref
7.
van den Broek  SP, Reinders  F, Donderwinkel  M, Peters  MJ.  Volume conduction effects in EEG and MEG.  Electroencephalogr Clin Neurophysiol. 1998;106(6):522-534.PubMedGoogle ScholarCrossref
8.
Stefan  H, Hummel  C, Scheler  G,  et al.  Magnetic brain source imaging of focal epileptic activity: a synopsis of 455 cases.  Brain. 2003;126(pt 11):2396-2405.PubMedGoogle ScholarCrossref
9.
Bartolomei  F, Trébuchon  A, Bonini  F,  et al.  What is the concordance between the seizure onset zone and the irritative zone? a SEEG quantified study.  Clin Neurophysiol. 2016;127(2):1157-1162.PubMedGoogle ScholarCrossref
10.
Asano  E, Juhász  C, Shah  A, Sood  S, Chugani  HT.  Role of subdural electrocorticography in prediction of long-term seizure outcome in epilepsy surgery.  Brain. 2009;132(pt 4):1038-1047.PubMedGoogle ScholarCrossref
11.
Stefan  H, Schneider  S, Feistel  H,  et al.  Ictal and interictal activity in partial epilepsy recorded with multichannel magnetoelectroencephalography: correlation of electroencephalography/electrocorticography, magnetic resonance imaging, single photon emission computed tomography, and positron emission tomography findings.  Epilepsia. 1992;33(5):874-887.PubMedGoogle ScholarCrossref
12.
Eliashiv  DS, Elsas  SM, Squires  K, Fried  I, Engel  J  Jr.  Ictal magnetic source imaging as a localizing tool in partial epilepsy.  Neurology. 2002;59(10):1600-1610.PubMedGoogle ScholarCrossref
13.
Medvedovsky  M, Taulu  S, Gaily  E,  et al.  Sensitivity and specificity of seizure-onset zone estimation by ictal magnetoencephalography.  Epilepsia. 2012;53(9):1649-1657.PubMedGoogle ScholarCrossref
14.
Assaf  BA, Karkar  KM, Laxer  KD,  et al.  Ictal magnetoencephalography in temporal and extratemporal lobe epilepsy.  Epilepsia. 2003;44(10):1320-1327.PubMedGoogle ScholarCrossref
15.
Shiraishi  H, Watanabe  Y, Watanabe  M, Inoue  Y, Fujiwara  T, Yagi  K.  Interictal and ictal magnetoencephalographic study in patients with medial frontal lobe epilepsy.  Epilepsia. 2001;42(7):875-882.PubMedGoogle ScholarCrossref
16.
Sutherling  WW, Crandall  PH, Engel  J  Jr, Darcey  TM, Cahan  LD, Barth  DS.  The magnetic field of complex partial seizures agrees with intracranial localizations.  Ann Neurol. 1987;21(6):548-558.PubMedGoogle ScholarCrossref
17.
Tang  L, Mantle  M, Ferrari  P,  et al.  Consistency of interictal and ictal onset localization using magnetoencephalography in patients with partial epilepsy.  J Neurosurg. 2003;98(4):837-845.PubMedGoogle ScholarCrossref
18.
Tilz  C, Hummel  C, Kettenmann  B, Stefan  H.  Ictal onset localization of epileptic seizures by magnetoencephalography.  Acta Neurol Scand. 2002;106(4):190-195.PubMedGoogle ScholarCrossref
19.
Yagyu  K, Takeuchi  F, Shiraishi  H,  et al.  The applications of time-frequency analyses to ictal magnetoencephalography in neocortical epilepsy.  Epilepsy Res. 2010;90(3):199-206.PubMedGoogle ScholarCrossref
20.
Perucca  P, Dubeau  F, Gotman  J.  Intracranial electroencephalographic seizure-onset patterns: effect of underlying pathology.  Brain. 2014;137(pt 1):183-196.PubMedGoogle ScholarCrossref
21.
Wetjen  NM, Marsh  WR, Meyer  FB,  et al.  Intracranial electroencephalography seizure onset patterns and surgical outcomes in nonlesional extratemporal epilepsy.  J Neurosurg. 2009;110(6):1147-1152.PubMedGoogle ScholarCrossref
22.
Bagic  A, Funke  ME, Ebersole  J; ACMEGS Position Statement Committee.  American Clinical MEG Society (ACMEGS) position statement: the value of magnetoencephalography (MEG)/magnetic source imaging (MSI) in noninvasive presurgical evaluation of patients with medically intractable localization-related epilepsy.  J Clin Neurophysiol. 2009;26(4):290-293.PubMedGoogle ScholarCrossref
23.
Foldvary  N, Klem  G, Hammel  J, Bingaman  W, Najm  I, Lüders  H.  The localizing value of ictal EEG in focal epilepsy.  Neurology. 2001;57(11):2022-2028.PubMedGoogle ScholarCrossref
24.
Alkawadri  R, Krishnan  B, Kakisaka  Y,  et al.  Localization of the ictal onset zone with MEG using minimum norm estimate of a narrow band at seizure onset versus standard single current dipole modeling.  Clin Neurophysiol. 2013;124(9):1915-1918.PubMedGoogle ScholarCrossref
25.
Hämäläinen  MS, Ilmoniemi  RJ.  Interpreting magnetic fields of the brain: minimum norm estimates.  Med Biol Eng Comput. 1994;32(1):35-42.PubMedGoogle ScholarCrossref
26.
Ioannides  AA, Bolton  JPR, Clarke  CJS.  Continuous probabilistic solutions to the biomagnetic inverse problem.  Inverse Probl. 1990;6(4):523-542.Google ScholarCrossref
27.
Uutela  K, Hämäläinen  M, Somersalo  E.  Visualization of magnetoencephalographic data using minimum current estimates.  Neuroimage. 1999;10(2):173-180.PubMedGoogle ScholarCrossref
28.
Wang  ZI, Jin  K, Kakisaka  Y,  et al.  Interconnections in superior temporal cortex revealed by musicogenic seizure propagation.  J Neurol. 2012;259(10):2251-2254.PubMedGoogle ScholarCrossref
29.
Tao  JX, Baldwin  M, Hawes-Ebersole  S, Ebersole  JS.  Cortical substrates of scalp EEG epileptiform discharges.  J Clin Neurophysiol. 2007;24(2):96-100.PubMedGoogle ScholarCrossref
30.
Barkley  GL.  Controversies in neurophysiology: MEG is superior to EEG in localization of interictal epileptiform activity: pro.  Clin Neurophysiol. 2004;115(5):1001-1009.PubMedGoogle ScholarCrossref
31.
Baillet  S, Mosher  JC, Leahy  RM.  Electromagnetic brain mapping.  IEEE Signal Process Mag. 2001;18(6):14-30. doi:10.1109/79.962275Google ScholarCrossref
32.
Gallen  CC, Sobel  DF, Waltz  T,  et al.  Noninvasive presurgical neuromagnetic mapping of somatosensory cortex.  Neurosurgery. 1993;33(2):260-268.PubMedGoogle ScholarCrossref
33.
Schneider  F, Alexopoulos  AV, Wang  Z,  et al.  Magnetic source imaging in non-lesional neocortical epilepsy: additional value and comparison with ICEEG.  Epilepsy Behav. 2012;24(2):234-240.PubMedGoogle ScholarCrossref
34.
Tenney  JR, Fujiwara  H, Horn  PS, Rose  DF.  Comparison of magnetic source estimation to intracranial EEG, resection area, and seizure outcome.  Epilepsia. 2014;55(11):1854-1863.PubMedGoogle ScholarCrossref
35.
Baumgartner  C, Pataraia  E, Lindinger  G, Deecke  L.  Neuromagnetic recordings in temporal lobe epilepsy.  J Clin Neurophysiol. 2000;17(2):177-189.PubMedGoogle ScholarCrossref
36.
Enatsu  R, Mikuni  N, Usui  K,  et al.  Usefulness of MEG magnetometer for spike detection in patients with mesial temporal epileptic focus.  Neuroimage. 2008;41(4):1206-1219.PubMedGoogle ScholarCrossref
37.
Pellegrino  G, Hedrich  T, Chowdhury  R,  et al.  Source localization of the seizure onset zone from ictal EEG/MEG data.  Hum Brain Mapp. 2016;37(7):2528-2546.PubMedGoogle ScholarCrossref
38.
Fujiwara  H, Greiner  HM, Hemasilpin  N,  et al.  Ictal MEG onset source localization compared to intracranial EEG and outcome: improved epilepsy presurgical evaluation in pediatrics.  Epilepsy Res. 2012;99(3):214-224.PubMedGoogle ScholarCrossref
39.
Knowlton  RC, Razdan  SN, Limdi  N,  et al.  Effect of epilepsy magnetic source imaging on intracranial electrode placement.  Ann Neurol. 2009;65(6):716-723.PubMedGoogle ScholarCrossref
40.
Almubarak  S, Alexopoulos  A, Von-Podewils  F,  et al.  The correlation of magnetoencephalography to intracranial EEG in localizing the epileptogenic zone: a study of the surgical resection outcome.  Epilepsy Res. 2014;108(9):1581-1590.PubMedGoogle ScholarCrossref
41.
Vadera  S, Jehi  L, Burgess  RC,  et al.  Correlation between magnetoencephalography-based “clusterectomy” and postoperative seizure freedom.  Neurosurg Focus. 2013;34(6):E9.PubMedGoogle ScholarCrossref
42.
Alkawadri  R, Burgess  R, Isitan  C, Wang  IZ, Kakisaka  Y, Alexopoulos  AV.  Yield of repeat routine MEG recordings in clinical practice.  Epilepsy Behav. 2013;27(2):416-419.PubMedGoogle ScholarCrossref
43.
Jeong  W, Kim  JS, Chung  CK.  Usefulness of multiple frequency band source localizations in ictal MEG.  Clin Neurophysiol. 2016;127(2):1049-1056.PubMedGoogle ScholarCrossref
44.
Kakisaka  Y, Alkawadri  R, Wang  ZI,  et al.  Sensitivity of scalp 10-20 EEG and magnetoencephalography.  Epileptic Disord. 2013;15(1):27-31.PubMedGoogle Scholar
Original Investigation
October 2018

Assessment of the Utility of Ictal Magnetoencephalography in the Localization of the Epileptic Seizure Onset Zone

Author Affiliations
  • 1The Epilepsy Center at Cleveland Clinic Foundation, Cleveland, Ohio
  • 2Yale Comprehensive Epilepsy Center, School of Medicine, Yale University, New Haven, Connecticut
  • 3Yale Human Brain Mapping Program, School of Medicine, Yale University, New Haven, Connecticut
  • 4The Department of Pediatrics, Tohoku University School of Medicine, Sendai, Japan
JAMA Neurol. 2018;75(10):1264-1272. doi:10.1001/jamaneurol.2018.1430
Key Points

Question  What is the chance of recording a seizure and the localizing yield of seizures recorded during routine magnetoencephalography (MEG) studies?

Findings  In this medical record analysis of 44 patients, MEG provided unique and more focal localization information, including in some electroencephalography and interictal MEG silent cases and in some cases that were nonlocalizable otherwise. Extended-source localization was more suitable for analysis of ictal rhythms than commonly used single-point solutions.

Meaning  If feasible, ictal MEG should be sought and considered in presurgical evaluation of drug-resistant epilepsy.

Abstract

Importance  Literature on ictal magnetoencephalography (MEG) in clinical practice and the relationship to other modalities is limited because of the brevity of routine studies.

Objective  To investigate the utility and reliability of ictal MEG in the localization of the epileptogenic zone.

Design, Setting, and Participants  A retrospective medical record review and prospective analysis of a novel ictal rhythm analysis method was conducted at a tertiary epilepsy center with a wide base of referrals for epilepsy surgery evaluation and included consecutive cases of patients who experienced epileptic seizures during routine MEG studies from March 2008 to February 2012. A total of 377 studies screened. Data were analyzed from November 2011 to October 2015.

Main Outcomes and Measures  Presurgical workup and interictal and ictal MEG data were reviewed. The localizing value of using extended-source localization of a narrow band identified visually at onset was analyzed.

Results  Of the 44 included patients, the mean (SD) age at the time of recording was 19.3 (14.9) years, and 25 (57%) were male. The mean duration of recording was 51.2 minutes. Seizures were provoked by known triggers in 3 patients and were spontaneous otherwise. Twenty-five patients (57%) had 1 seizure, 6 (14%) had 2, and 13 (30%) had 3 or more. Magnetoencephalography single equivalent current dipole analysis was possible in 29 patients (66%), of whom 8 (28%) had no clear interictal discharges. Sublobar concordance between ictal and interictal dipoles was seen in 18 of 21 patients (86%). Three patients (7%) showed clear ictal MEG patterns without electroencephalography changes. Ictal MEG dipoles correlated with the lobe of onset in 7 of 8 patients (88%) who underwent intracranial electroencephalography evaluations. Reasons for failure to identify ictal dipoles included diffuse or poor dipolar ictal patterns, no MEG changes, and movement artifact. Resection of areas containing a minimum-norm estimate of a narrow band at onset, not single equivalent current dipole, was associated with sustained seizure freedom.

Conclusions and Significance  Ictal MEG data can provide reliable localization, including in cases that are difficult to localize by other modalities. These findings support the use of extended-source localization for seizures recorded during MEG.

Introduction

Magnetoencephalography (MEG) is increasingly used in specialized epilepsy centers for the evaluation of patients with intractable epilepsy. Magnetoencephalography presents several advantages, including its noninvasiveness, high temporal and spatial resolution, and ability to provide a broad whole-head view of brain activities. In many studies, MEG has been demonstrated to add nonredundant information to guide implantation sites for intracranial recordings,1,2 resulting in changes to surgical treatment and medical management.3 Magnetoencephalography has some advantages over routine electroencephalography (EEG) and other imaging modalities4; the magnetic field is less affected by the conductivity of the intervening tissue, and fast-frequency activity, which is more relevant for seizure localization, is less susceptible than scalp EEG to muscle activity.5-7

Magnetoencephalography is routinely used to analyze the irritative zone. The use of interictal discharges, although practical, is conceptually challenging. First, MEG’s sensitivity to interictal discharges is 70%.8 Second, it is common to encounter multiple areas of active interictal spiking that do not necessarily overlap with the seizure onset zone.9 This is in accordance with previous studies that found that recording of the seizure onset zone remains the best marker of the epileptogenic zone.10 Because of logistical considerations, recording of seizures may not be feasible during routine studies. As a consequence, ictal MEG studies have reported on up to 26 patients only,11-13 and in the largest cohort to date, recording lasted up to 40 hours, much longer than routine studies.13 There is no consensus concerning the best way to process these magnetic ictal data. Single equivalent current dipole (sECD) modeling was used in most studies.11-18 However, ictal discharges are dynamic, and a single-point model, while practical, is conceptually challenging. Furthermore, seizures may start as rhythmic waves or fast activity, with an amplitude much lower than that of spikes. The lower amplitude activity may be difficult to localize on each time point considered independently with an sECD.19 However, the fast, low-amplitude frequencies are more specific for the localization of the epileptogenic zone based on intracranial EEG data.20,21 In the largest cohort on the subject to date, we report our experience at a tertiary epilepsy center. We also examine the efficiency of a semiautomated method implementing extended-source localization of a narrow band at the onset.

Methods
Patients

From a total of 377 consecutive MEG studies performed at the Cleveland Clinic from March 2008 to February 2012, we identified patients who experienced epileptic seizures during their studies. The protocol was approved by the Cleveland Clinic Institutional Review Board. No study-specific consent was required for this retrospective medical record analysis because standard data deidentification and cross-referencing was implemented.

The analysis of sECD was performed for clinical purposes, and the results were presented in the surgical management conference. Magnetoencephalography was typically performed in patients who were medically intractable and had a variable degree of discordance in the presurgical workup or in those whose EEGs were not informative. The decision about surgical plan was based on expert group consensus in a weekly surgical conference where the results of MEG were discussed. The extended-source localization of a narrow band was performed prospectively in a blinded manner. Patients were instructed to report auras or spells. Also, it was routinely inquired if seizures/auras were experienced after the study. We included patients who experienced epileptic auras or magnetic seizures. We excluded patients with nonepileptic paroxysmal events and patients with MEG studies with inadequate quality.

MEG Data Acquisition and Analysis

Approximately 1 hour of MEG activity was recorded using the whole-head 306-channel Elekta Neuromag Vector View system. Electroencephalography was recorded simultaneously on the same machine (21 channels; 10–20 system). The files were divided into 10-minute segments. Data were sampled at 1000 Hz and filtered with 0.1 Hz and 300 Hz. Evaluation of the sources producing the interictal activity of interest was carried out in Elekta Neuromag Data Analysis Software, DANA 3.3, Xfit source-modeling software, version 5.5.18, using multiple-sphere head model. Dipoles were coregistered with the patient’s magnetic resonance imaging (MRI) using routine fiducial points. Sensors were used to identify the position of the patient’s head to compensate for movements during the procedure. Source modeling was based on MEG data only. We defined the ictal onset as the first evolving rhythmic oscillations temporally related to the clinical and EEG onset of the patient’s typical seizures. Both EEG and MEG data were reviewed by the MEG team, which included 2 expert epileptologists active in the American Clinical Magnetoencephalography Society.22

The analysis was performed on data segments that contained interictal and ictal onset discharges using standard modeling parameters. The dipoles were estimated from a subset of 30 to 60 channels around the maximum peak (Figure 1) and were coregistered on the patient’s MRI using a MEG-MRI coordinate integration. The number of channels chosen depended on the extent of the field and was implemented to enhance the localization of the signal of interest and minimize fitting interferences from distant unrelated activity. Dipole fits were deemed acceptable when pretest modeling parameters were met (goodness of fit, >85%; confidence volume, <1 cm3). The onset and rhythms were identified as the earliest evolving changes in the background, as described by Foldvary et al.23 The location, orientation, and strength of the dipole sources that best fit the measured magnetic fields were calculated with Elekta Neuromag Data Analysis Software using least squares algorithms. For analysis of ictal rhythms, stability of the magnetic influx and efflux was verified visually by expert readers (Figure 1), the earliest ictal discharge was used for dipole fitting, and dipoles were accepted if they met the appropriate fitting criteria (Figure 1). For rhythms other than spikes, we fit the dipoles at the time of the peak of waveforms. Although generalized discharges rendered dipole fits on occasions, it is conceptually challenging to accept single-point solutions for such widespread activity unless there is a clear electrical/magnetic lead, which should be the subject of dipole fitting then.

Rhythm/Frequency Band–Based Method of Analysis of Ictal Rhythms of the Minimum-Norm Estimate of a Narrow Band

For details of the method, refer to the technical report by Alkawadri et al24 and an L2 minimum-norm estimation of sources (Figure 2; eAppendix in the Supplement).25-27 Briefly, a band centered around the dominant frequency at seizure onset is determined based on averaged spectra from visually identified sensors at the time of onset. The L2-MNE is applied on the band of interest and coregistered with an MRI-based cortical brain mesh.

Statistical Analysis

We used standard measures for descriptive statistics and Fisher exact test to examine the correlation of the resection of the ictal MEG areas with surgical outcomes. We used t test statistics for comparisons of means. The hypotheses for statistical comparisons were determined prehoc. No exploratory statistical analyses were performed. Significance was set at a P value less than .05, and all P values were 2-tailed unless stated otherwise.

Results
Demographic and Clinical Characteristics of Patients

Forty-four of 377 patients (11.7%) were included (Table). Nineteen patients (43%) were female. Eighteen patients (41%) had normal findings on MRI of the brain prior to the study, 22 (50%) had MRI signal abnormalities (eg, cavernous malformations, focal cortical dysplasias, trauma), and 4 (9%) had prior resections along with no other apparent lesion. Most of the patients were classified by video EEG criteria as having generalized seizures, bilateral nonlocalizable seizures, or independent bilateral seizures. Only 17 patients (39%) were classified as having exclusively regional seizure onset by scalp video EEG, reflecting the complexity of referral base. Seizures were provoked by known stimuli in 3 patients (7%) and were spontaneous in 41 patients (93%). The average duration of MEG study in included patients was 51.2 minutes and the mean (SD) age at the time of the recording was 19.3 (14.9) years, whereas the mean (SD) age of excluded patients who underwent routine MEG study during the same period was 26.0 (16.4) years (P = .01). Baseline seizure frequency as reported by patients was 182 seizures per month, whereas the average seizure frequency in excluded patients was 98.7 seizures per month (P < .001).

Ictal MEG Findings

Twenty-five patients (57%) had 1 seizure during routine MEG recording, 6 (14%) had 2 seizures, 7 (16%) had 3 to 20 seizures, and 5 (11%) had more than 20 seizures. There was 1 patient who was referred for persistent focal seizures and arrived at the MEG laboratory for localization of the status. Seizures were induced in 3 patients using known triggers (eg, sensory or music); details of the musicogenic case were reported by Wang et al.28 Unlike scalp EEG, MEG tended to show more focal onsets, which was regional or unilateral in 26 patients (59%), some generalized and some localized onsets in 4 (9%), some generalized or nonlocalizable in 8 (18%), obscured by movement in 3 (7%), and no MEG change in 3 (7%). Equivalent sECD was successful in all recoded seizures in 25 patients (57%), successful in some but not all seizures in 4 patients (9%), and unsuccessful in 15 patients (34%).

MEG and Concurrent Scalp EEG

Magnetoencephalography provided unique localizing findings in 16 patients (36%) not seen on simultaneous scalp EEG. This included 3 patients (7%) with simple partial seizures with MEG changes and no EEG changes, 6 (14%) with nonlocalizable or generalized EEG changes, and 4 (9%) with discordant localization compared with EEG. In all 4, MEG lobar and sublobar localization was confirmed by subsequent intracranial EEG studies.

In most patients, the type of onset was similar in MEG and EEG. There were 3 patients (7%) with auras that showed ictal patterns in MEG and no EEG changes. On the other hand, there were no patients that showed EEG changes without MEG changes. In 6 patients (14%), MEG showed onsets as paroxysmal fast or repetitive spiking, whereas EEG showed less localizing, slower rhythms. In most patients, EEG and MEG onsets were concurrent; MEG preceded EEG onset in 4 patients (9%) and EEG preceded MEG in 1 (2%).

Comparison of Localization
Ictal MEG vs Interictal Dipole Fitting

Dipole fitting was feasible in 35 patients (80%) with interictal onset discharge and 29 patients (66%) with ictal onset discharge. There were 8 patients (18%) in whom ictal MEG provided unique findings not detected interictally, of whom 7 (16%) showed no MRI abnormalities and 5 (11%) showed nonlocalizing findings on video EEG. Of these, localization was concordant with invasive EEG data in 3 patients (7%), consensus of surgical conference based on multimodality workup in 3 (7%), and video EEG in 1 (2%). The localization remained uncertain in 1 patient in whom MEG showed localizing findings. Lobar concordance between the most frequent interictal9 and the ictal dipole fits was seen in all 21 patients in whom both were available. Of these, sublobar concordance was seen in 18 patients (86%).

Ictal MEG Dipole Fit vs Patient Management Conference or Video EEG Conclusion

Epilepsy was localized or lateralized in 31 patients (70%) and was nonlocalizable in 13 (30%) based on the consensus of experts in the surgical management conference or video EEG findings when conference consensus was not available. Ictal MEG provided consistent localizing data in 5 of 13 patients (38%). On the other hand, ictal MEG analysis was concordant with the localization determined by other modalities in 24 of 31 patients (77%).

Ictal MEG Dipole Fit vs Onset Intracranial EEG Data

Eleven patients (25%) underwent invasive evaluations. In 8, the seizure onset was localizable on intracranial EEG and ictal sECD analysis was feasible. Of these, lobar and sublobar concordance with ictal MEG dipole were seen in 7 (88%) and 6 (75%) patients, respectively.

Results of Narrow Band Minimum-Norm Estimate Analysis

Sixteen patients with onsets characterized as paroxysmal fast or poor dipolar patterns were included in the analysis (Figure 2). Concordance of localization was seen in 7 of 8 patients where the results of analysis were available in both, ie, sECD fell within the region identified by rhythm-based method. Of the 12 patients with known surgical outcomes, minimum-norm estimate of a narrow band (MNE-fc) was feasible in 9 (75%). Six patients (14%) remained seizure free, all 44 patients had ictal MNE-fc regions resected, and in the 3 patients with recurrence of seizures, the resection of the ictal MNE-fc areas was not complete. In 1 patient (Figure 2F) in whom discordance between sECD and MNE-fc was seen, the patient became seizure free, although the sECD area was not resected whereas the MNE-fc areas were resected.24 In this patient, dipole analysis was not feasible. The surgical resection of areas identified by MNE of a narrow band correlated with good surgical outcomes.

Correlation With Surgical Outcomes

Sixteen patients (36%) underwent resection of the suspected focus and were followed up after more than 2 years (range, 26-62 months). Resection of the area that contained the dipoles did not correlate with outcomes (risk ratio, 2.5; 95% CI, 0.76-8.19; 1-tailed Fisher exact test, P = .12). The complete resection of areas delineated by MNE-fc correlated with seizure freedom (risk ratio, 3.5; 95% CI, 1.08-11.9; 1-tailed Fisher exact test, P = .01).

Discussion

In this study, we demonstrated that ictal events can be recorded in the MEG suite in a substantial number of routine 1-hour studies compared with other practices using longer durations.13 Our cohort represented 12% of patients with refractory epilepsy who were referred to our tertiary surgical center.

Ictal sECD analysis was feasible in 66% of patients despite the conceptual challenges underlying the use of the model in evolving sources at seizure onset. Furthermore, ictal MEG provided unique findings in one-third of patients who had seizures that were difficult to localize otherwise, including video EEG–negative cases despite the relative brevity of the recordings. Magnetoencephalography also showed changes in 3 patients that were not seen on concurrent EEG (50% of patients with negative concurrent EEG results and 7% of total cohort). Of interest, 2 of these patients had prior resections, highlighting perhaps MEG’s known lower susceptibility to the conductivity of the intervening tissue and structural changes. We speculate that this finding is independent of the underlying pathology. Besides reporting on the incidence of seizures visualized on MEG but not EEG, no firm conclusions could be made because of the small number of patients in this subgroup. There was high concordance between interictal and ictal MEG dipoles when both were present. Magnetoencephalography superiority in localizing the epileptic focus does stem in part from the greater number of the sensors compared with scalp EEG. However, other factors play a role; the size of spiking cortex (5 × 5 mm) needed for MEG detection is smaller than the area of the cortex for scalp EEG detection.29-31 This may account for the earlier and higher chance of detection of seizures compared with standard EEG as well as the fact that analysis was limited by movement artifact only in 3 patients. We chose to report on lobar and sublobar concordance rather than geometrical distance for several reasons. It is known that errors that relate to coregistration alone account for about 4 to 7 millimeters for procedures that are done correctly4,32 and may be higher in some studies accounting for other causes of variability. These values are smaller than the interelectrode distance in the standard intracranial EEG sampling, often used as the criterion standard for localization.33,34

Ictal MEG poses some methodological challenges. The patient must have seizures during the MEG recording, the head movement needs to be compensated, and movement-related magnetic artifacts must be suppressed. Moreover, at ictal onset, both EEG and MEG signals may have a low signal-to-noise ratio, which may not always allow for accurate dipole analysis. Additionally, the magnetic field attenuates rapidly as the distance from the focus to MEG sensors increases, and distant sources in the deeply seated and mesial temporal cortex foci might go undetected.35,36

Minimum-norm estimate–based source localization of a narrow band at the onset has theoretical advantages over single-point solutions when tackling dynamic and evolving rhythms. We demonstrated the practicality and effectiveness of this method. The frequency band is specific to the seizure and patient. This method provided additional and more reliable localizing information in 3 of 12 patients with surgical outcome data and adequate follow-up after resective surgery and was concordant with dipole analysis when both were feasible otherwise. Furthermore, it was concordant with the lobar localization based on surgical conference discussion in 4 patients in whom no surgical procedure took place. We demonstrate that the resection of the areas delineated by the MNE-fc method had a stronger correlation with seizure freedom than single-point analysis. Our findings, along with the ones from a 2016 study,37 support the use of extended-source localization of seizure-specific frequency bands as a primary method for analysis of seizures recorded during MEG. It is important to investigate the performance of different inverse problem solutions for localization of the ictal rhythms, including beamforming techniques. A high degree of agreement is expected, similar to those reported previously for interictal discharges recognizing the limitations associated with analysis of the latter for localization of the epileptic focus.34 However, our method of band-specific and seizure-specific extended-source localization does not make a priori assumptions other than restricting sources to the cortex and is more suitable in theory and practically to evaluate dynamic sources, such as those encountered at seizure onset, compared with point solutions. Future studies may elucidate the reliability of beamforming methods for analysis of ictal rhythms.

To our knowledge, this is the largest series on the localizing yield of ictal MEG. It highlights the importance of recording MEG seizures, if possible, especially in difficult-to-localize cases. Differences in reporting and methods may account for differences of sensitivity with available studies.13 Concordance between interictal MEG and invasive onset is comparable with results reported by Eliashiv et al.12 The concordance between interictal and ictal MEG is consistent with the study by Tang et al.17 Our results support the high concordance with intracranial EEG reported in a few studies.13,38 It has been our approach to integrate MEG in our presurgical planning, with an emphasis on patients with discordance in the localization of presurgical evaluation and, in particular, instances with nonlocalizable or paucity of localizing EEG findings. We and others have previously demonstrated the additive localizing value of interictal MEG analysis independent of other modalities3,39,40 as well as the added benefit of clusteroctomy and correlation with better surgical outcomes.41 We have also demonstrated the additional yield of repeated MEG studies, which is comparable with the yield of repeated routine EEG.42 The findings of the current study extend present knowledge and highlight the unique and additional localizing information provided by ictal MEG. Ictal MEG recordings should be sought whenever possible, and consideration could be made to arrange for MEG recordings while patients are in the monitoring units on reduced doses of antiseizure medications if feasible, provided the MEG suite is staffed suitably.

Limitations and Future Directions

There are inherent limitations related to the retrospective design. The results are applicable to this cohort, and recording seizures may vary because of differences in complexity of cases. Other factors may influence outcomes that were not controlled for. Future studies may elucidate the reliability of analysis of predetermined bands (eg, β or high-frequency oscillations).43 We deferred analysis in this study as some of the patients had significant aliasing from electromagnetic artifact requiring multinotch filtering, which in turns limits the reliability of analysis in a number of patients. Also, a study to evaluate time needed to record seizures will address questions regarding cost-effectiveness of prolonged recordings. Among other limitations related to ictal MEG, foremost, it is an inherently difficult test to perform. Magnetoencephalography should be interpreted by a clinical magnetoencephalographer. Furthermore, despite the high concordance with intracranial EEG in this cohort, MEG is not yet poised to replace intracranial EEGs as it is prone to errors of coregistration, can only detect cortical sources when they exceed certain area thresholds, and can be challenging with deeply seated foci. The goal of this study was to report on practical aspects. In theory, comparison of yields of MEG and EEG as a modality per se should be performed using the same number of sensors. The latter is challenging for routine practices in EEG and is less safe and convenient for patients. In that regard, and in addition to our findings, it has been shown that MEG is less susceptible to the conductivity of intervening tissue and that the size of sources of detectable MEG signal is much smaller than that for EEG.44

Conclusions

Ictal MEG may provide novel information in a considerable number of presurgical evaluations, even in some EEG silent cases or cases that are difficult to localize otherwise. Extended-source localization of a narrow band at ictal onset should be considered as the primary method of analysis, especially when dipole fitting is problematic. Resection of the areas delineated by extended-source localization of a narrow band at onset correlates with a higher chance of seizure freedom. Future studies may elucidate the effectiveness of source localization of predetermined frequency bands, such as β and high-frequency oscillations.

Back to top
Article Information

Accepted for Publication: April 9, 2018.

Corresponding Author: Rafeed Alkawadri, MD, Yale Human Brain Mapping Program, School of Medicine, Yale University, 15 York St, LCI 714B, New Haven, CT 06510 (mhdrafeed.alkawadri@yale.edu).

Published Online: June 11, 2018. doi:10.1001/jamaneurol.2018.1430

Author Contributions: Dr Alkawadri had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Alkawadri, Burgess, Kakisaka, Alexopoulos.

Acquisition, analysis, or interpretation of data: Alkawadri, Burgess, Mosher, Alexopoulos.

Drafting of the manuscript: Burgess.

Critical revision of the manuscript for important intellectual content: Alkawadri, Kakisaka, Mosher, Alexopoulos.

Statistical analysis: Mosher.

Administrative, technical, or material support: Alkawadri, Burgess, Mosher.

Study supervision: Alkawadri, Burgess, Alexopoulos.

Conflict of Interest Disclosures: None reported.

Funding/Support: This publication was made possible by funding support from grant 412064 from the American Epilepsy Society and Clinical and Translational Science Award grant UL1 TR000142 from the National Center for Advancing Translational Science.

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The contents of this article are solely the responsibility of the authors and do not necessarily represent the official view of National Institutes of Health.

Additional Contributions: We acknowledge the generous support of the Swebilius trust. We thank magnetoencephalography technologists at the Cleveland Clinic, Cleveland, Ohio, for their help in data acquisition.

References
1.
Knowlton  RC, Elgavish  R, Howell  J,  et al.  Magnetic source imaging versus intracranial electroencephalogram in epilepsy surgery: a prospective study.  Ann Neurol. 2006;59(5):835-842.PubMedGoogle ScholarCrossref
2.
Agirre-Arrizubieta  Z, Thai  NJ, Valentín  A,  et al.  The value of magnetoencephalography to guide electrode implantation in epilepsy.  Brain Topogr. 2014;27(1):197-207.PubMedGoogle ScholarCrossref
3.
Sutherling  WW, Mamelak  AN, Thyerlei  D,  et al.  Influence of magnetic source imaging for planning intracranial EEG in epilepsy.  Neurology. 2008;71(13):990-996.PubMedGoogle ScholarCrossref
4.
Knowlton  RC, Laxer  KD, Aminoff  MJ, Roberts  TP, Wong  ST, Rowley  HA.  Magnetoencephalography in partial epilepsy: clinical yield and localization accuracy.  Ann Neurol. 1997;42(4):622-631.PubMedGoogle ScholarCrossref
5.
Zimmermann  R, Scharein  E.  MEG and EEG show different sensitivity to myogenic artifacts.  Neurol Clin Neurophysiol. 2004;2004:78.PubMedGoogle Scholar
6.
Claus  S, Velis  D, Lopes da Silva  FH, Viergever  MA, Kalitzin  S.  High frequency spectral components after secobarbital: the contribution of muscular origin–a study with MEG/EEG.  Epilepsy Res. 2012;100(1-2):132-141.PubMedGoogle ScholarCrossref
7.
van den Broek  SP, Reinders  F, Donderwinkel  M, Peters  MJ.  Volume conduction effects in EEG and MEG.  Electroencephalogr Clin Neurophysiol. 1998;106(6):522-534.PubMedGoogle ScholarCrossref
8.
Stefan  H, Hummel  C, Scheler  G,  et al.  Magnetic brain source imaging of focal epileptic activity: a synopsis of 455 cases.  Brain. 2003;126(pt 11):2396-2405.PubMedGoogle ScholarCrossref
9.
Bartolomei  F, Trébuchon  A, Bonini  F,  et al.  What is the concordance between the seizure onset zone and the irritative zone? a SEEG quantified study.  Clin Neurophysiol. 2016;127(2):1157-1162.PubMedGoogle ScholarCrossref
10.
Asano  E, Juhász  C, Shah  A, Sood  S, Chugani  HT.  Role of subdural electrocorticography in prediction of long-term seizure outcome in epilepsy surgery.  Brain. 2009;132(pt 4):1038-1047.PubMedGoogle ScholarCrossref
11.
Stefan  H, Schneider  S, Feistel  H,  et al.  Ictal and interictal activity in partial epilepsy recorded with multichannel magnetoelectroencephalography: correlation of electroencephalography/electrocorticography, magnetic resonance imaging, single photon emission computed tomography, and positron emission tomography findings.  Epilepsia. 1992;33(5):874-887.PubMedGoogle ScholarCrossref
12.
Eliashiv  DS, Elsas  SM, Squires  K, Fried  I, Engel  J  Jr.  Ictal magnetic source imaging as a localizing tool in partial epilepsy.  Neurology. 2002;59(10):1600-1610.PubMedGoogle ScholarCrossref
13.
Medvedovsky  M, Taulu  S, Gaily  E,  et al.  Sensitivity and specificity of seizure-onset zone estimation by ictal magnetoencephalography.  Epilepsia. 2012;53(9):1649-1657.PubMedGoogle ScholarCrossref
14.
Assaf  BA, Karkar  KM, Laxer  KD,  et al.  Ictal magnetoencephalography in temporal and extratemporal lobe epilepsy.  Epilepsia. 2003;44(10):1320-1327.PubMedGoogle ScholarCrossref
15.
Shiraishi  H, Watanabe  Y, Watanabe  M, Inoue  Y, Fujiwara  T, Yagi  K.  Interictal and ictal magnetoencephalographic study in patients with medial frontal lobe epilepsy.  Epilepsia. 2001;42(7):875-882.PubMedGoogle ScholarCrossref
16.
Sutherling  WW, Crandall  PH, Engel  J  Jr, Darcey  TM, Cahan  LD, Barth  DS.  The magnetic field of complex partial seizures agrees with intracranial localizations.  Ann Neurol. 1987;21(6):548-558.PubMedGoogle ScholarCrossref
17.
Tang  L, Mantle  M, Ferrari  P,  et al.  Consistency of interictal and ictal onset localization using magnetoencephalography in patients with partial epilepsy.  J Neurosurg. 2003;98(4):837-845.PubMedGoogle ScholarCrossref
18.
Tilz  C, Hummel  C, Kettenmann  B, Stefan  H.  Ictal onset localization of epileptic seizures by magnetoencephalography.  Acta Neurol Scand. 2002;106(4):190-195.PubMedGoogle ScholarCrossref
19.
Yagyu  K, Takeuchi  F, Shiraishi  H,  et al.  The applications of time-frequency analyses to ictal magnetoencephalography in neocortical epilepsy.  Epilepsy Res. 2010;90(3):199-206.PubMedGoogle ScholarCrossref
20.
Perucca  P, Dubeau  F, Gotman  J.  Intracranial electroencephalographic seizure-onset patterns: effect of underlying pathology.  Brain. 2014;137(pt 1):183-196.PubMedGoogle ScholarCrossref
21.
Wetjen  NM, Marsh  WR, Meyer  FB,  et al.  Intracranial electroencephalography seizure onset patterns and surgical outcomes in nonlesional extratemporal epilepsy.  J Neurosurg. 2009;110(6):1147-1152.PubMedGoogle ScholarCrossref
22.
Bagic  A, Funke  ME, Ebersole  J; ACMEGS Position Statement Committee.  American Clinical MEG Society (ACMEGS) position statement: the value of magnetoencephalography (MEG)/magnetic source imaging (MSI) in noninvasive presurgical evaluation of patients with medically intractable localization-related epilepsy.  J Clin Neurophysiol. 2009;26(4):290-293.PubMedGoogle ScholarCrossref
23.
Foldvary  N, Klem  G, Hammel  J, Bingaman  W, Najm  I, Lüders  H.  The localizing value of ictal EEG in focal epilepsy.  Neurology. 2001;57(11):2022-2028.PubMedGoogle ScholarCrossref
24.
Alkawadri  R, Krishnan  B, Kakisaka  Y,  et al.  Localization of the ictal onset zone with MEG using minimum norm estimate of a narrow band at seizure onset versus standard single current dipole modeling.  Clin Neurophysiol. 2013;124(9):1915-1918.PubMedGoogle ScholarCrossref
25.
Hämäläinen  MS, Ilmoniemi  RJ.  Interpreting magnetic fields of the brain: minimum norm estimates.  Med Biol Eng Comput. 1994;32(1):35-42.PubMedGoogle ScholarCrossref
26.
Ioannides  AA, Bolton  JPR, Clarke  CJS.  Continuous probabilistic solutions to the biomagnetic inverse problem.  Inverse Probl. 1990;6(4):523-542.Google ScholarCrossref
27.
Uutela  K, Hämäläinen  M, Somersalo  E.  Visualization of magnetoencephalographic data using minimum current estimates.  Neuroimage. 1999;10(2):173-180.PubMedGoogle ScholarCrossref
28.
Wang  ZI, Jin  K, Kakisaka  Y,  et al.  Interconnections in superior temporal cortex revealed by musicogenic seizure propagation.  J Neurol. 2012;259(10):2251-2254.PubMedGoogle ScholarCrossref
29.
Tao  JX, Baldwin  M, Hawes-Ebersole  S, Ebersole  JS.  Cortical substrates of scalp EEG epileptiform discharges.  J Clin Neurophysiol. 2007;24(2):96-100.PubMedGoogle ScholarCrossref
30.
Barkley  GL.  Controversies in neurophysiology: MEG is superior to EEG in localization of interictal epileptiform activity: pro.  Clin Neurophysiol. 2004;115(5):1001-1009.PubMedGoogle ScholarCrossref
31.
Baillet  S, Mosher  JC, Leahy  RM.  Electromagnetic brain mapping.  IEEE Signal Process Mag. 2001;18(6):14-30. doi:10.1109/79.962275Google ScholarCrossref
32.
Gallen  CC, Sobel  DF, Waltz  T,  et al.  Noninvasive presurgical neuromagnetic mapping of somatosensory cortex.  Neurosurgery. 1993;33(2):260-268.PubMedGoogle ScholarCrossref
33.
Schneider  F, Alexopoulos  AV, Wang  Z,  et al.  Magnetic source imaging in non-lesional neocortical epilepsy: additional value and comparison with ICEEG.  Epilepsy Behav. 2012;24(2):234-240.PubMedGoogle ScholarCrossref
34.
Tenney  JR, Fujiwara  H, Horn  PS, Rose  DF.  Comparison of magnetic source estimation to intracranial EEG, resection area, and seizure outcome.  Epilepsia. 2014;55(11):1854-1863.PubMedGoogle ScholarCrossref
35.
Baumgartner  C, Pataraia  E, Lindinger  G, Deecke  L.  Neuromagnetic recordings in temporal lobe epilepsy.  J Clin Neurophysiol. 2000;17(2):177-189.PubMedGoogle ScholarCrossref
36.
Enatsu  R, Mikuni  N, Usui  K,  et al.  Usefulness of MEG magnetometer for spike detection in patients with mesial temporal epileptic focus.  Neuroimage. 2008;41(4):1206-1219.PubMedGoogle ScholarCrossref
37.
Pellegrino  G, Hedrich  T, Chowdhury  R,  et al.  Source localization of the seizure onset zone from ictal EEG/MEG data.  Hum Brain Mapp. 2016;37(7):2528-2546.PubMedGoogle ScholarCrossref
38.
Fujiwara  H, Greiner  HM, Hemasilpin  N,  et al.  Ictal MEG onset source localization compared to intracranial EEG and outcome: improved epilepsy presurgical evaluation in pediatrics.  Epilepsy Res. 2012;99(3):214-224.PubMedGoogle ScholarCrossref
39.
Knowlton  RC, Razdan  SN, Limdi  N,  et al.  Effect of epilepsy magnetic source imaging on intracranial electrode placement.  Ann Neurol. 2009;65(6):716-723.PubMedGoogle ScholarCrossref
40.
Almubarak  S, Alexopoulos  A, Von-Podewils  F,  et al.  The correlation of magnetoencephalography to intracranial EEG in localizing the epileptogenic zone: a study of the surgical resection outcome.  Epilepsy Res. 2014;108(9):1581-1590.PubMedGoogle ScholarCrossref
41.
Vadera  S, Jehi  L, Burgess  RC,  et al.  Correlation between magnetoencephalography-based “clusterectomy” and postoperative seizure freedom.  Neurosurg Focus. 2013;34(6):E9.PubMedGoogle ScholarCrossref
42.
Alkawadri  R, Burgess  R, Isitan  C, Wang  IZ, Kakisaka  Y, Alexopoulos  AV.  Yield of repeat routine MEG recordings in clinical practice.  Epilepsy Behav. 2013;27(2):416-419.PubMedGoogle ScholarCrossref
43.
Jeong  W, Kim  JS, Chung  CK.  Usefulness of multiple frequency band source localizations in ictal MEG.  Clin Neurophysiol. 2016;127(2):1049-1056.PubMedGoogle ScholarCrossref
44.
Kakisaka  Y, Alkawadri  R, Wang  ZI,  et al.  Sensitivity of scalp 10-20 EEG and magnetoencephalography.  Epileptic Disord. 2013;15(1):27-31.PubMedGoogle Scholar
×