ARTICLE OPEN ACCESS CLASS OF EVIDENCE

Electromagnetic source imaging in presurgicalworkup of patients with epilepsyA prospective study

Lene Duez MD PhD Hatice Tankisi MD PhD Peter Orm Hansen MD PhD Per Sidenius MD DMSc

Anne Sabers MD DMSc Lars H Pinborg MD DMSc Martin Fabricius MD DMSc Gyorgy Rasonyi MD

Guido Rubboli MD DMSc Birthe Pedersen MD Anne-Mette Leffers MD Peter Uldall MD DMSc

Bo Jespersen MD Jannick Brennum MD DMSc Otto Moslashlby Henriksen MD PhD

Anders Fuglsang-Frederiksen MD DMSc and Sandor Beniczky MD PhD

Neurologyreg 201992e576-e586 doi101212WNL0000000000006877

Correspondence

Prof Beniczky

sandorbeniczky

aarhusrmdk

AbstractObjectiveTo determine the diagnostic accuracy and clinical utility of electromagnetic source imaging(EMSI) in presurgical evaluation of patients with epilepsy

MethodsWe prospectively recorded magnetoencephalography (MEG) simultaneously with EEG andperformed EMSI comprising electric source imaging magnetic source imaging and analysis ofcombined MEG-EEG datasets using 2 different software packages As reference standard forirritative zone (IZ) and seizure onset zone (SOZ) we used intracranial recordings and forlocalization accuracy outcome 1 year after operation

ResultsWe included 141 consecutive patients EMSI showed localized epileptiform discharges in 94patients (67) Most of the epileptiform discharge clusters (72) were identified by bothmodalities 15 only by EEG and 14 only by MEG Agreement was substantial betweeninverse solutions and moderate between software packages EMSI provided new informationthat changed the management plan in 34 of the patients and these changes were useful in80 Depending on the method EMSI had a concordance of 53 to 89 with IZ and 35 to73 with SOZ Localization accuracy of EMSI was between 44 and 57 which was notsignificantly different fromMRI (49ndash76) and PET (54ndash85) Combined EMSI achievedsignificantly higher odds ratio compared to electric source imaging and magnetic sourceimaging

ConclusionEMSI has accuracy similar to established imaging methods and provides clinically useful newinformation in 34 of the patients

Classification of evidenceThis study provides Class IV evidence that EMSI had a concordance of 53ndash89 and35ndash73 (depending on analysis) for the localization of epileptic focus as compared withintracranial recordingsmdashIZ and SOZ respectively

MORE ONLINE

CME CourseNPuborgcmelist

Class of EvidenceCriteria for ratingtherapeutic and diagnosticstudies

NPuborgcoe

These authors contributed equally to this work

From the Departments of Clinical Neurophysiology (LD HT POH AF-F SB) and Neurology (PS) Aarhus University Hospital Departments of Neurology (AS LHP) ClinicalNeurophysiology (MF G Rasonyi) Pediatrics Child Neurology (PU) Neurosurgery (BJ JB) and Clinical Physiology Nuclear Medicine and PET (OMH) Copenhagen UniversityHospital Rigshospitalet Danish Epilepsy Centre (G Rubboli BP SB) Dianalund and Department of Diagnostic Radiology (A-ML) Hvidovre Hospital Denmark

Go to NeurologyorgN for full disclosures Funding information and disclosures deemed relevant by the authors if any are provided at the end of the article

The Article Processing Charge was funded by the Danish Epilepsy Centre Filadelfia

This is an open access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License 40 (CC BY-NC-ND) which permits downloadingand sharing the work provided it is properly cited The work cannot be changed in any way or used commercially without permission from the journal

e576 Copyright copy 2019 The Author(s) Published by Wolters Kluwer Health Inc on behalf of the American Academy of Neurology

Epilepsy has a prevalence of 4ndash101000 persons and up to30 do not respond to antiepileptic drugs1 Surgery isa therapeutic option for patients with drug-resistant focalepilepsy2 Depending on the ability to localize and resect theepileptogenic zone 30 to 85 of operated patients becomeseizure-free3

The greatest challenge in epilepsy surgery is the presurgicalevaluation On their own none of the methods can reliablyidentify the epileptogenic zone Therefore a multimodal ap-proach using semiology EEG and MRI is needed These areoften supplemented with functional neuroimaging methodssuch as SPECT or PET4 In difficult cases intracranial elec-trodes are implanted3

Advances in recording techniques and in signal analysis havemade it possible to estimate the source of the epileptiformdischarges (EDs) recorded by EEG (EEG source imaging ESI)and magnetoencephalography (MEG source imaging MSI)4

In studies56 investigating the clinical utility of MSI (ie changesin clinical decision-making based onMSI) MSI led to a changein the management plan in 21 to 33 of patients and this wasassociated with long-term seizure control Large prospectivestudies showed that ESI accurately estimated the location of theepileptic focus78 although its role in clinical decision-makingwas not addressed in those studies

It is possible to record MEG and EEG signals simultaneouslyand perform electromagnetic source imaging (EMSI) analyzingboth modalities either separately (ESI and MSI) or combined(cEMSI) Despite numerous studies suggesting that thesemethods are accurate ESI and MSI are only implemented ina few centers5 and often EEG and MEG are not recordedsimultaneously This is probably attributable to the lack of well-designed large prospective blinded studies addressing EMSIMany questions remain unanswered What is the diagnosticyield of simultaneousMEG and EEG recordingsWhich inversesolution is best Do different software packages and operatorsgive similar results Can the methods accurately estimate thesource of the EDs Do they provide nonredundant clinicallyuseful information that changes and improves decision-makingin presurgical evaluation

To elucidate the role of EMSI in presurgical evaluation weconducted a prospective blinded study We evaluated the

diagnostic yield of the 2 recorded modalities inter-softwareand -operator agreement accuracy of source localization andits clinical utility We designed the study and we reported itaccording to the STARD (Standards for Reporting DiagnosticAccuracy) criteria9

MethodsStudy design and patientsFrom April 2011 to June 2016 we prospectively recruitedconsecutive patients undergoing presurgical evaluation aspart of the Danish Epilepsy Surgery Program All patients haddrug-resistant focal epilepsy and for all patients at least 2appropriately chosen and taken antiepileptic drugs had failedThe intended sample size was ge100 patients based on pre-vious studies10 This number has proven to provide the ref-erence standard in gt50 patients undergoing intracranialrecording (ICR) or operation11

Inclusion criteria were patients who underwent presurgicalevaluation and who gave their consent There were no ex-clusion criteria data from all recruited patients were analyzed

Standard protocol approvals registrationsand patient consentsThe study was approved by the regional ethics committee(case number 1-10-72-177-12) and patients gave their in-formed written consent Patients underwent presurgicalevaluation according to the investigation protocol approvedby the Danish Health Authority (data available from Dryadappendix 1 doiorg105061dryadp4r01pq) This includeddetailed history semiology and EEG from video-EEG mon-itoring MRI and neuropsychological tests for all patientsWhen decisions could not be made based on these datapatients were referred to additional noninvasive investigations(PET EMSI) and ICRs (stereo-EEG)

Presurgical patients in whom the multidisciplinary team(MDT) reached a conclusion based on semiology EEG andMRI were recruited as part of the research study (arm A) Inthese patients EMSI was not part of the decision-makingPatients with normal MRI or discordant data (semiologyEEG MRI) were referred to PET and EMSI as part of theirpresurgical evaluation before ICR (arm B) In these patientsEMSI was part of the decision-making process

GlossaryBESA = Brain Electrical Source Analysis cEMSI = combined electromagnetic source imaging DSM = distributed sourcemodel ECD = equivalent current dipole ED = epileptiform discharge EMSI = electromagnetic source imaging ESI = EEGsource imaging FDG = 18F-fluorodeoxyglucose FLAIR = fluid-attenuated inversion recovery FN = false negative FP = falsepositive GLM = generalized linear model ICR = intracranial recording IZ = irritative zone MDT = multidisciplinary teamMEG = magnetoencephalography MPRAGE = magnetization-prepared rapid-acquisition gradient echo MSI = magneto-encephalography source imaging NPV = negative predictive value OR = odds ratio PPV = positive predictive value SOZ =seizure onset zone TN = true negative TP = true positive

NeurologyorgN Neurology | Volume 92 Number 6 | February 5 2019 e577

For patients in arm B theMDT had a 2-step decision processDecision was first made blinded to EMSI based on all otherdata and it was logged into the database Decisions werecategorized as (1) stop (operation not offered) (2) implan-tation of intracranial electrodes (specifying their location) or(3) operation (resective surgery) Then a second decisionwas made based on new information (if any) provided by theEMSI weighted against all other data Possible changes in themanagement plan included any change from one of thesecategories to another and change within category 2(ie change in the planned intracranial electrode positions)

Simultaneous MEG and EEG recordingsMEG-EEG was recorded at Aarhus University Hospital usingan Elekta Neuromag system (Elekta Stockholm Sweden)

MEG was recorded at 102 evenly distributed locations with 3sensors at each location (2 orthogonal planar gradiometersand a magnetometer) giving a total of 306 channels Con-tinuous head position indicators were used MEG data werepreprocessed offline using the spatiotemporal signal spaceseparation method12 to suppress the residual interference andto correct for head movements

High-density EEG (64ndash80 electrodes) was recorded usinga nonmagnetic cap (EASYCAP) and additional electrodescovering the inferior part of the head (cheek inferior temporalchain and neck) in all patients with head circumference lt60cm and who could cooperate with the procedure In otherpatients an array of 19 electrodes (10-20 system) or no EEGwas recorded In addition electrooculogram and ECG wererecorded Electronic tracking of positions of electrodes and ofthe head position coils were done using a Polhemus system(Polhemus Colchester VT)13

Spontaneous activity (eyes closed rest supine position) wasrecorded for 60 to 90 minutes (sampling frequency 1 kHzonline bandpass 01ndash330 Hz) both for MEG and EEG Datawere offline bandpass filtered at 05 to 70 Hz

Electromagnetic source imagingMEG-EEG was inspected for EDs by 2 trained experts (LDSB) Signals were analyzed using 2 commercially availablesoftware packages with European approval for medical use(CE-mark) LD used the CURRY 7 Neuroimaging Suite(Compumedics Victoria Australia) SB used BESA-Research 6 (BESA Grafelfing Germany) Voltage maps andmagnetic field maps were inspected and EDs were grouped indifferent clusters when the topographic maps differed Thenfor each cluster using the visually detected EDs as templatesautomated algorithms (template matching) scanned therecordings and detected EDs were visually checked To im-prove the signal-to-noise ratio EDs with similar topographywere averaged14

Sequential topographic maps were inspected in the time ep-och of the ascending slope of the averaged waveforms for

change in topography indicating propagation In additionprincipal components analysis was used to identify propaga-tion15 When propagation occurred source imaging was donefirst at the onset epoch and then at the peak When propa-gation was not identified source imaging was done at themiddle third of the ascending slope10 With both softwarepackages 2 different inverse solutions were applied equiva-lent current dipole (ECD) and a distributed source model(DSM) sLORETA16 with an anatomically constrained linearestimation approach assuming the sources were distributed inthe cerebral cortex in CURRY and CLARA in BESA8 ESI andMSI were done with both software packages In additionusing CURRY source modeling of both datasets (EEG-MEG) was done (cEMSI)17

All analyses were performed using individual head modelsfrom the patientsrsquo MRIs coregistered with the MEG-EEGrecording using fiducial landmark positions electrode posi-tions and head position coils digitized before the recordingIn CURRY 3-compartment boundary element method (BEM)head model was used18 In BESA an individual 4-layer finiteelement method (FEM) head model was used14 Figure 1illustrates the steps of the EMSI

EMSI was performed prospectively blinded to clinical in-formation and reference standard results

Conventional neuroimaging methodsMRI and 18F-fluorodeoxyglucose (FDG)-PET were per-formed as part of the presurgical evaluation The MRIepilepsy protocol consisted until 2015 of a sagittal T13-dimensional (3D) magnetization-prepared rapid-acquisitiongradient echo (MPRAGE) sequence coronal and axial fluid-attenuated inversion recovery (FLAIR) sequences anda coronal T2-weighted (W) fast spin-echo sequence Cor-onal and axial FLAIR and T2W sequences were performedperpendicular to and in plane with hippocampus long-axisSince 2015 the protocol consisted of sagittal T1 3DMPRAGE sequence sagittal 3D FLAIR sequence coronaland axial T2W sequences perpendicular and horizontal tohippocampus long-axis axial susceptibility-weighted imagesequence axial diffusion-weighted image sequence sagittalT1 3D MPRAGE sequence and axial T1W spin-echo se-quence after IV contrast All 3D sequences were recon-structed in coronal and axial planes perpendicular andhorizontal to the hippocampus long-axis

For PET FDG was administered IV with the patient insupine position and equipped with eye pads and earplugsPET scan was commenced 40 minutes after tracer in-jection PET scans were coregistered and displayedsuperimposed on MRI sequences using a standard clinicalworkstation and dedicated software supplied by the cam-era vendor (Leonardo station using TrueD and from 2013syngovia using MI Neurology Siemens Medical SolutionsMalvern PA) who also provided access to age-matchednormal database comparison

e578 Neurology | Volume 92 Number 6 | February 5 2019 NeurologyorgN

Sublobar brain regionsEMSI PET andMRI localization results were prospectivelylogged into the database with localization scores at sublobarlevel corresponding to 64 categories This included thesublobar regions proposed by an international consensuspaper and IFCN (International Federation of ClinicalNeurophysiology) guideline1519 A descriptor of the side(leftrightmesial) was added When more than 2 in-dependent foci were identified and none of them hada dominant activity the result was scored as multifocal Afocus was considered dominant if the number of EDs wasmore than twice the number of EDs from all other foci Theremaining categories were (1) normal and (2) not localiz-able Not localizable was when the ECD confidence interval

or the distributed source analysis spanned over more thanone lobe

Reference standardIdeal reference standard lacks for source imaging in presur-gical evaluation6 Accurately localizing the generator ofinterictal EDs (the irritative zone [IZ]) does not necessarilyimply that resection of that area will render the patientseizure-free Therefore we used 2 different sets of referencestandards ICR and outcome after surgery (see below)

Outcome measuresFirst the number of ED clusters recorded by only MEG andonly EEG were compared

Figure 1 Methodologic flowchart of EMSI

EMSI was performed using the following 4 steps (1)review of the magnetoencephalography and EEGrecordings and visual identification and marking ofEDs belonging to the same cluster These EDs wereused for template-matching Each detected spike wasvisually checked and artifacts were discarded (2) EDswithin each cluster were averaged to improve signal-to-noise ratio (critical for spike onset activity relativeto the background activity)28 Sequential topographicplots of the ascending phase and principal compo-nents analysis was used to identify propagation (3)Individual head model was created for each patientand the EEG electrodes were aligned to the scalp (4)Source modeling was performed using 2 differentinverse-solution strategies equivalent current dipoleand distributed source models where yellow indi-cates maximum intensity ED = epileptiform dis-charge EMSI = electromagnetic source imaging IED =interictal epileptiform discharge

NeurologyorgN Neurology | Volume 92 Number 6 | February 5 2019 e579

Agreement between different EMSI methods was calculatedAll comparisons were done at sublobar level

Clinical utility of EMSI was defined as the proportion ofpatients in whom EMSI changed the decision of the MDT Achange was defined useful as follows (1) change from stop toICRmdashthe ICR localized the source (2) change in implanta-tion strategymdashthe electrode(s) implanted based on the EMSIidentified the source (3) change from implantation to op-eration without preceding implantationmdashthe patient becameseizure-free (at 1-year postoperative follow-up)

For EMSI PET and MRI we calculated the following out-come measures Agreement with ICR concordance with IZand concordance with the seizure onset zone (SOZ) weredetermined using ICR as reference standard Proportion ofpatients with results concordant with the reference standardand agreement with this reference standard were calculated

Sensitivity specificity localization accuracy positive pre-dictive value (PPV) and negative predictive value (NPV)were determined using outcome 1 year after surgery as ref-erence standard Preoperative EMSI was coregistered with thepostoperative MRI the source was considered concordantwith the resected site when the ECD and respectively themaximum of the DSM was within the operation cavity Allother results (localization outside the cavity multifocal EDsnormal recoding not-localizable results) were considereddiscordant We categorized the localizations as follows TP(true positive)mdashconcordant with the resection and seizure-free FP (false positive)mdashconcordant with the resection andnot seizure-free TN (true negative)mdashdiscordant with theresection and not seizure-free FN (false negative)mdashdiscordant with the resection and seizure-free We used thefollowing formula

Sensitivity = TP=ethTP+ FNTHORN

Specificity = TN=ethTN+FPTHORN

Accuracy = ethTN+TPTHORN=ethTP + FP +TN+FNTHORN

PPV = TP=ethTP+ FPTHORN

NPV = TN=ethTN+FNTHORNSurgical procedures were tailored to each patient based onthe final conclusion of the MDT

We calculated the odds ratio (OR) of becoming seizure-freewhen operation was concordant vs discordant with the lo-calization of the index test OR = (TPFP)(FNTN)

StatisticsFor assessment of agreement (κ) between EMSI methodssoftware packages and operators and for assessment ofagreement between EMSI and ICR we calculated GwetAC1

20 We used this measure of agreement because it yields

more reasonable values than the Cohen κ in case of low-traitprevalences20 Interrater agreement was interpreted accordingto the conventional groups poor (κ lt 0) slight (κ 001ndash02)fair (κ 021ndash04) moderate (κ 041ndash06) substantial (κ061ndash08) and almost perfect agreement (κ gt 08)21

The effect measures of sensitivity specificity PPV NPV con-cordance with SOZ localization accuracy and their 95 con-fidence intervals were estimated using a generalized linearmodel (GLM)2223 Identity link function was used in the GLMto compare index tests by means of the differences of the effectmeasure Multiple observations per patient were considered inthe GLM by specifying the patient ID as a cluster Analysis wasperformed in StataCorp software release 14 (StataCorp LPCollege Station TX) using the binreg function

Classification of evidenceThe study was designed to answer the following questionWhat is the localization accuracy and clinical utility of EMSI inpresurgical evaluation of patients with drug-resistant focalepilepsy

The study was prospective and blinded All patients whounderwent presurgical evaluation and who gave their in-formed consent were recruited Data from all recruitedpatients were analyzed and evaluated All patients had drug-resistant focal epilepsy

This study provides Class IV evidence that EMSI had a con-cordance of 53ndash89 and 35ndash73 (depending on analy-sis) for the localization of epilepsy as compared withICRsmdashIZ and SOZ respectively

Data sharingIndividual deidentified data including localization results ofthe imaging methods reference standard localizations andoutcome data will be shared along with the following relateddocuments study protocol and statistical analysis plan Un-restricted access to these data will be made available from theday of the online publication of the article until 2030 usinga publicly accessible repository (datadryadorg) (Dryaddoi105061dryadp4r01pq) During the review phase thedata package is temporarily available via the following linkdatadryadorgreviewdoi=doi105061dryadp4r01pq

ResultsFigure 2 shows examples of EMSI Source images coregisteredwith postoperative MRIs are shown in data available fromDryad (appendix 2 doiorg105061dryadp4r01pq)Figure 3 shows the flowchart of presurgical evaluation of allrecruited patients

PatientsOne hundred forty-one consecutive patients (59 females)were included Median age was 32 years (range 8ndash70) All

e580 Neurology | Volume 92 Number 6 | February 5 2019 NeurologyorgN

patients had MRI and MEG Simultaneously with MEG 115patients had high-density EEG (64ndash80 electrodes) Becauseof large head circumference or inability to cooperate 8patients had an array of 19 EEG electrodes and 18 patients didnot have EEG One hundred ten patients had PET MRIshowed potentially epileptogenic lesions in 68 patients (48)PET showed a functional deficit zone in 70 patients (63)Seventy-two patients had temporal foci and 66 had extra-temporal foci (frontal 50 parietal 11 occipital 5) Threepatients could not be localized The time from MEG-EEGrecording to operation ranged from 1 to 48 months No ad-verse events occurred from performing MEG-EEG

Detection of EDs by EEG and MEGEMSI showed focus localized to one or more sublobar regionsin 94 patients (67) The remaining patients had normalMEG-EEG (46 patients 33) or the source was not local-izable (one patient) See data available from Dryad appen-dices 3 and 4 doiorg105061dryadp4r01pq

In patients who had simultaneous MEG and EEG recordings(n = 123) a total of 96 clusters of EDs were identified in 85patients (60 of all patients) In 72 (7096) EDs werevisible both in MEG and EEG In 15 (1496) EDs werevisible only in EEG In 13 (1296) EDs were visible only inMEG There was no significant difference in the number ofinterictal foci detected only by EEG and only by MEG (p =067) Of the 14 foci detected only by EEG 9 were located inthe temporal lobe Of the 12 foci detected only by MEG 7were temporal

Agreement between EMSI methodsBetween-software and -operator agreement both for ECDand for DSM agreement between BESA and CURRY wasmoderate for both modalities (ESI or MSI) (table 1)

Agreement in localization between the 2 inverse solutionstrategies (ECD and DSM) within the same software andoperator using BESA there was almost perfect agreementfor MSI and substantial agreement for ESI (table 1) Whenusing CURRY there was substantial agreement betweenECD and DSM for both modalities (ESI and MSI)(table 1)

Clinical utility of EMSIThe effect of EMSI on patient management planning wasassessed in 85 patients (50 males age 10ndash70 median 32years) in whom EMSI was part of the decision-making pro-cess In this subgroup 38 patients were MRI-negative in theremaining patients there were discordances between MRIand data from long-term video-EEG monitoring (semiologyand EEG)

In 34 (2985) EMSI changed the management planFigure 4 shows the types of changes in the management planbased on EMSI At 1-year follow-up reference standard wasavailable for 20 patients (5 patients did not wish to proceedwith ICR one patient withdrew before operation and 3patients were still on the waiting list) In 80 (1620) thesechanges proved to be useful in 44 patients who proceeded toimplantation based on EMSI IZ and SOZ had been identi-fied in 810 patients in whom additional electrodes wereimplanted the additional electrodes identified IZ and SOZ46 patients who skipped implantation and proceeded tooperation became seizure-free See data available from Dryadappendix 5 doiorg105061dryadp4r01pq

Concordance with ICRFifty-three patients (34 males age 8ndash60 median 40 years)underwent ICR 25 were MRI-negative Twenty-nine patientshad temporal foci and 24 had extratemporal foci (frontal 16

Figure 2 Electromagnetic source imaging

Electromagnetic source imaging (equivalent current dipole and distributed source model) for a patient with frontal (A and B) and temporal (C and D) focusAnalysis was done using CURRY (A and C) and BESA (B and D) software Figures in appendix 2 (data available from Dryad doiorg105061dryadp4r01pq)show preoperative sources coregistered with postoperative MRI for these patients

NeurologyorgN Neurology | Volume 92 Number 6 | February 5 2019 e581

parietal 6 occipital 2) SOZ could not be localized by ICR in14 of 53 patients

Concordance between EMSI and the IZ was between 53and 89 (table 2) For ECD this was higher in BESA than inCURRY both for EEG (p = 00003) and MEG (p = 00107)for DSM there was no significant difference between the 2software packages Concordance with IZ was higher for mostof the EMSI methods than for MRI (50 31ndash69) and forPET (29 13ndash45) (data available from Dryad appen-dices 6 and 7 doiorg105061dryadp4r01pq)

Agreement between BESA EMSI and IZ was substantialAgreement between CURRY MEG DSM and IZ was

substantial for the other CURRY EMSI it was moderate(table 3) Agreement with IZ was moderate for MRI and fairfor PET (table 3)

Concordance with SOZ for the EMSI methods was between35 and 73 (table 2) For ECD this was higher in BESAthan in CURRY both for EEG (p = 0003) and MEG (p =0002) For DSM there was no difference between the 2software packages Concordance with SOZ for CURRY DSMcEMSI and CURRY ECD MEG was significantly higher thanforMRI (p = 0036 and p = 0029 respectively) (data availablefrom Dryad appendix 6 doiorg105061dryadp4r01pq)Concordance with SOZ for CURRY ECD MEG was higherthan for PET (p = 0033) Concordance with SOZ for the

Table 1 Agreement between electromagnetic sourceimaging methods

MEG EEG

BESA vs CURRY

ECD 058 (048ndash069) 047 (034ndash058)

DSM 058 (049ndash068) 043 (033ndash054)

BESA

ECD vs DSM 085 (078ndash091) 072 (063ndash082)

CURRY

ECD vs DSM 071 (062ndash080) 063 (052ndash073)

Abbreviations DSM = distributed source model ECD = equivalent currentdipole MEG = magnetoencephalographyBetween-software and -operator agreement (BESA vs CURRY) and agree-ment between inverse solution strategies (ECD vs DSM) within the samesoftware and operator Data represent κ values (confidence intervals)

Figure 4 Clinical utility of EMSI

In 34 of patients (2985) EMSI changed the management plan Thechanges were distributed as follows stop rarr implantation of intracranialelectrodes 685 (7) implantationrarr stop 185 (1) change in the locationof implanted electrodes 1485 (165) skipping implantation and goingdirectly to operation 885 (94) EMSI = electromagnetic source imaging IC= intracranial electrodes

Figure 3 Flowchart of the presurgical evaluation for the 141recruited patients

Red arrows and boxes indicate that operation was not offered and greenarrows andboxes indicate that operationwas indicated by theMDT At thisstage in the flowchart the MDTmade 2-step decisions first blinded to EMSIthen including EMSI results One patient died of acute myocardial in-farction and one patient died of sudden unexpected death in epilepsy EMSI= electromagnetic source imaging ICR = intracranial recording MDT =multidisciplinary team OP = operation

e582 Neurology | Volume 92 Number 6 | February 5 2019 NeurologyorgN

Table 2 Performance of electromagnetic source imaging MRI and PET in presurgical evaluation

Reference standard ICR Seizure-free patients 1 y after operation

Outcome measure Concordance with IZ Concordance with SOZ Sensitivity Specificity PPV NPV Localization accuracy Odds ratio

CURRY ECD EEG 56 (37ndash74) 35 (14ndash56) 10 (0ndash22) 90 (78ndash100) 60 (17ndash100) 42 (28ndash57) 44 (30ndash58) 13 (02ndash72)

CURRY DSM EEG 73 (56ndash90) 50 (28ndash72) 17 (3ndash31) 90 (78ndash100) 71 (38ndash100) 44 (29ndash59) 48 (34ndash62) 08 (01ndash56)

CURRY ECD MEG 65 (48ndash81) 35 (15ndash55) 33 (16ndash50) 90 (78ndash100) 83 (62ndash100) 49 (33ndash65) 57 (43ndash71) 47 (07ndash308)

CURRY DSM MEG 84 (72ndash96) 67 (48ndash86) 43 (25ndash61) 76 (58ndash95) 72 (51ndash93) 48 (31ndash66) 57 (43ndash71) 20 (04ndash103)

CURRY ECD cEMSI 53 (31ndash71) 36 (16ndash67) 31 (14ndash48) 81 (64ndash98) 69 (44ndash95) 46 (30ndash62) 52 (38ndash66) 58 (09ndash364)

CURRY DSM cEMSI 80 (66ndash94) 64 (43ndash84) 38 (20ndash56) 76 (58ndash95) 69 (46ndash92) 47 (30ndash64) 54 (40ndash68) 192 (18ndash2047)

BESA ECD EEG 87 (75ndash99) 73 (54ndash92) 31 (14ndash48) 71 (52ndash91) 60 (35ndash85) 43 (26ndash59) 48 (34ndash62) 10 (02ndash52)

BESA DSM EEG 74 (59ndash90) 64 (43ndash84) 34 (17ndash52) 71 (52ndash91) 63 (39ndash86) 44 (27ndash61) 50 (36ndash64) 05 (01ndash32)

BESA ECD MEG 89 (78ndash99) 65 (45ndash85) 43 (25ndash61) 76 (58ndash95) 72 (51ndash93) 48 (31ndash66) 57 (43ndash71) 30 (06ndash151)

BESA DSM MEG 82 (69ndash95) 63 (43ndash82) 43 (25ndash61) 76 (58ndash95) 72 (51ndash93) 48 (31ndash66) 57 (43ndash71) 36 (07ndash174)

MRI 50 (31ndash69) 37 (15ndash89) 67 (50ndash84) 57 (36ndash79) 69 (52ndash86) 55 (34ndash76) 63 (49ndash76) mdash

PET 29 (13ndash45) 39 (16ndash62) 59 (35ndash82) 81 (62ndash100) 77 (54ndash100) 65 (44ndash86) 70 (54ndash86) mdash

Abbreviations cEMSI = combined electromagnetic source imaging DSM = distributed source model ECD = equivalent current dipole ICR = intracranial recording IZ = irritative zone MEG = magnetoencephalography NPV =negative predictive value PPV = positive predictive value SOZ = seizure onset zoneData represent proportion (95 confidence interval) Since none of the seizure-free patients were operated discordant to MRI or PET odds ratio could not be determined for these methods

Neurolo

gyorgN

Neurology

|Volum

e92N

umber

6|

February

52019e583

remaining EMSI methods did not differ from PET (dataavailable from Dryad appendix 7) Agreement between SOZand DSM cEMSI in CURRY and between SOZ and bothMSImethods in BESA (ECD and DSM) was moderate Theremaining EMSI analysis MRI and PET had a fair agreementwith SOZ (table 3)

Operated patientsOf the 57 patients who had resective surgery 51 patients (27males age 8ndash62 median 62 years) were operated more than1 year ago (figure 3) Thirty-one patients had temporal fociand 20 had extratemporal foci (frontal 15 parietal 3 occipital2) The median time from MEG-EEG recording to operationwas 10 months (range 1ndash48 months) Thirty patients (59)were seizure-free at 1-year postoperative follow-up (figure 3)

Table 2 summarizes the measures determined from the op-eration outcome sensitivity specificity PPV NPV and lo-calization accuracy See data available fromDryad appendices6 and 7 doiorg105061dryadp4r01pq

Localization accuracy (proportion of TPs and TNs) of EMSImethods was between 44 and 57 There was no significantdifference in localization accuracy among the EMSI methodsand between the EMSI methods and MRI There was nosignificant difference between the localization accuracy ofPET and EMSI in BESA and cEMSI in CURRY

Sensitivity of EMSI methods was between 10 and 43(table 2) For ESI using ECD sensitivity was significantly

higher in BESA than in CURRY (p = 002) There was nosignificant difference in sensitivity among the other EMSImethods Sensitivity of MRI (67 50ndash84) was signifi-cantly higher thanmost EMSImethods except forMSI (ECDand DSM) using BESA and DSM of MEG-EEG signals usingCURRY There was no significant difference in sensitivitybetween PET (59 35ndash82) and DSMs (cEMSI inCURRY MEG in CURRY and BESA EEG in BESA) andbetween PET and ECD ESI in BESA PET had significantlyhigher sensitivity than the other EMSI methods See dataavailable from Dryad appendices 6 and 7 doiorg105061dryadp4r01pq

Specificity of EMSI methods was between 71 and 90(table 2) ESI in BESA had significantly higher specificity thanusing CURRY Specificity was higher for all EMSI methodsthan MRI (57 36ndash79) and this was statistically signif-icant for all methods except for cEMSI There was no sig-nificant difference in specificity between EMSI methods andPET See data available from Dryad appendices 6 and 7 doiorg105061dryadp4r01pq

EMSI had PPV between 60 and 83 and NPV between42 and 49 (table 2) There was no significant differencein PPV among the EMSI methods and between the EMSImethods and conventional neuroimaging methods (MRIand PET) In addition there was no significant differencebetween NPV of MRI and EMSI NPV of PET was higherthan that of EMSI This was significant compared to ESI andMSI in CURRY the cEMSI analysis (ECD) in CURRY andfor the analysis of EEG using ECD in BESA See dataavailable from Dryad appendices 6 and 7 doiorg105061dryadp4r01pq

Since none of the seizure-free patients were operated dis-cordant to MRI or PET OR could not be determined forthese methods There was no significant difference in OR ofESI and MSI between the 2 software packages OR for MSIwas between 2 and 47 and for ESI from 05 to 125 (dataavailable from Dryad appendix 8 doiorg105061dryadp4r01pq) Combining both EEG and MEG signals (cEMSI)gave an OR of 58 for ECD and 192 for DSM (table 2) OR ofcEMSI using DSMwas significantly higher than both ESI (p =002) and MSI (p = 003)

Additional data with the distribution of the results in differentlesion types and locations (temporal vs extratemporal) areavailable from Dryad (appendix 9 doiorg105061dryadp4r01pq)

DiscussionMore than one-fourth (28) of the ED clusters were visible inonly one modality (MEG 13 EEG 15) Thus althoughmost clusters were visible in both modalities in order to re-cord all clusters simultaneous MEG and EEG recording is

Table 3 Agreement between imaging methods andintracranial recording

Agreement with IZ Agreement with SOZ

CURRY ECD cEMSI 058 (038ndash078) 038 (020ndash058)

CURRY DSM cEMSI 056 (038ndash073) 040 (022ndash058)

CURRY ECD MEG 049 (032ndash067) 033 (017ndash050)

CURRY DSM MEG 060 (043ndash077) 039 (024ndash055)

CURRY ECD EEG 058 (038ndash077) 027 (010ndash045)

CURRY DSM EEG 056 (038ndash073) 030 (014ndash049)

BESA ECD MEG 063 (047ndash080) 042 (024ndash058)

BESA DSM MEG 061 (045ndash078) 041 (025ndash060)

BESA ECD EEG 071 (052ndash087) 033 (016ndash051)

BESA DSM EEG 071 (051ndash088) 034 (015ndash055)

MRI 041 (013ndash069) 032 (006ndash056)

PET 032 (011ndash057) 032 (012ndash057)

Abbreviations cEMSI = combined electromagnetic source imaging DSM =distributed source model ECD = equivalent current dipole IZ = irritativezone MEG = magnetoencephalography SOZ = seizure onset zoneData represent κ values (95 confidence intervals)

e584 Neurology | Volume 92 Number 6 | February 5 2019 NeurologyorgN

needed EMSI showed focal epileptiform abnormalities lo-calized at sublobar level in 67 of the patients

Despite using standardized methodology variability in-troduced by different software packages and operators washigher than the variability introduced by using different in-verse solutions (ECD vs DSM) by the same operator in thesame software This calls for further standardization

EMSI methods achieved substantial agreement with IZbut only moderate with SOZ (MSI and cEMSI) This is notsurprising since EMSI analyzed interictal EDs which isthe IZ Although IZ is often in the same region as SOZICR clearly showed24 that this is not always the casePrevious studies suggested that the ldquodominantrdquo cluster ofinterictal discharges is concordant with the SOZ7 How-ever these studies did not define what was ldquodominantrdquoAlthough we have prospectively defined what we consid-ered dominant clusters we found that source imaging ofinterictal discharges correlated better with IZ thanwith SOZ

Using postoperative outcome as reference standard we foundthat the accuracy of EMSI was not significantly different fromMRI (none of the EMSI methods) and from PET (MSI inBESA and cEMSI in CURRY) Although sensitivity of MRIwas higher than most EMSI methods its specificity was lowerthan EMSI There was no significant difference in sensitivityand specificity between most EMSI methods and PET UsingDSM for analyzing the combined dataset of MEG and EEGyielded significantly higher OR than separate analysis of bothdatasets This further emphasized the clinical importance ofrecording MEG and EEG simultaneously

The accuracy of EMSI was relatively low though similar to theconventional neuroimaging methods Localizing the epilep-togenic zone is extremely difficult and there is no singlemethod that on its own is sufficient for the presurgical eval-uation Although in the whole group of patients the accuracyof EMSI was not significantly different from conventionalneuroimaging these methods complemented each othersince EMSI was able to localize the source in cases in whichthe conventional methods did not EMSI provided new in-formation that changed patientsrsquo management plan in one-third of the individual cases In 80 these changes proved tobe useful at 1-year follow-up This demonstrates the role ofEMSI in the multimodal presurgical evaluation of patientswith drug-resistant focal epilepsy

This study has several limitations Almost all patients in whomEMSI was part of the clinical workup consented to the study(normal MRI or discordant MRI EEG and semiology)However this was not the case for patients whose decisionabout operation was reached based on concordant MRI EEGand semiology and where EMSI was not part of the clinicalworkup Thus more complex cases might be overrepresentedin our study

All patients were undergoing a presurgical evaluation in theDanish national epilepsy surgery program and all MEG-EEGwere recorded and analyzed at Aarhus University HospitalAlthough we addressed the variability introduced by 2 dif-ferent software packages and analyzers our study is single-center

Perfect reference standard lacks for presurgical evaluation6

Because of spatial sampling problems ICRs can be mis-leading However using resection site and postoperativeoutcome as reference standard has its drawbacks too despitecorrect localization of IZ and SOZ patients might not becomeseizure-free since IZ and SOZ do not necessarily coincidewith the epileptogenic zone Therefore we opted for usingboth datasets as reference standard and here we emphasizethe intrinsic limitations of both approaches

Our high-density EEG array recorded simultaneously with theMEG was between 64 and 80 electrodes Although somestudies suggested that at least 128 electrodes are needed foroptimal ESI7 other studies failed to find increase in accuracybeyond 64 electrodes25 However not only the number ofelectrodes matters but also their distribution (even spatialsampling) and coverage of inferior parts of the head (cheekand neck) Our electrode array was in accordance with theseprinciples

Previous studies using simulated data somatosensoryresponses and a case study emphasized the importance ofindividual realistic conductivity estimations for combinedEEG and MEG source analysis172627 Because of the com-plementary nature of the 2 modalities and the increasednumber of sensors superior spatial resolution was achievedThis is in line with our findings that the OR was highest forcEMSI

The evidence provided by this study is classified as Class IVbecause implantation of intracranial electrodes was not blin-ded to the results of the EMSI and lt80 of the patients hadbeen implanted or operated Since locations shown by EMSIneed to be implanted for validation it is technically impossibleto do this blinded to the EMSI Furthermore typically lessthan half of the patients entering presurgical evaluation areimplanted or operated Therefore it is practically impossibleto achieve a higher class of evidence for diagnostic studies inpresurgical evaluation of patients with epilepsy

Our results indicate that EMSI is as accurate as the well-established neuroimaging methods and it provides clinicallyuseful nonredundant information for the presurgical evalua-tion of patients with epilepsy

Author contributionsLD AF-F and SB contributed to the conception anddesign of the study LD SB POH PS AS LHP MFG Rasonyi G Rubboli BP A-ML OMH PU BO andJB acquired and analyzed data LD and SB drafted the

NeurologyorgN Neurology | Volume 92 Number 6 | February 5 2019 e585

manuscript and the figures All authors contributed to editingthe final manuscript

AcknowledgmentThe authors express their gratitude to neurophysiologytechnicians Henrik Soslashvsoslash and Esben Holch Porsborg (AarhusUniversity Hospital) for the MEG-EEG recordings toChristopher Bailey (Aarhus University) for technical assis-tance with the MEG recordings to Michael Scherg (BESA)and Jorn Kastner (Compumedics) for assistance withdevelopment of the standardized analysis pipeline and toAparna Udupi (Aarhus University) for help with the statisticalanalysis Per Sidenius is a deceased author

Study fundingSupported by Aarhus University Lundbeck Foundation TheVelux Foundations the Danish Epilepsy Association Heleneand Viggo Bruuns Fund and Lennart Grams Memorial Fund

DisclosureL Duez H Tankisi and P Hansen report no disclosuresrelevant to the manuscript P Sidenius is deceased dis-closures are not included for this author A Sabers L PinborgM Fabricius G Rasonyi G Rubboli B Pedersen A LeffersP Uldall B Jespersen J Brennum O Henriksen and AFuglsang-Frederiksen report no disclosures relevant to themanuscript S Beniczky reports nonfinancial support fromElekta personal fees from Sage Therapeutics Brain SentinelEisai and UCB outside the submitted work Go to Neurol-ogyorgN for full disclosures

Publication historyReceived byNeurology April 11 2018 Accepted in final form October 22018

References1 Sander JW The epidemiology of epilepsy revisited Curr Opin Neurol 200316

165ndash1702 Kwan P Brodie MJ Early identification of refractory epilepsy N Engl J Med 2000

342314ndash3193 Rosenow F Luders H Presurgical evaluation of epilepsy Brain 20011241683ndash17004 Mouthaan BE Rados M Barsi P et al Current use of imaging and electromagnetic

source localization procedures in epilepsy surgery centers across Europe Epilepsia201657770ndash776

5 Knowlton RC Razdan SN Limdi N et al Effect of epilepsy magnetic source imagingon intracranial electrode placement Ann Neurol 200965716ndash723

6 De Tiege X Carrette E Legros B et al Clinical added value of magnetic sourceimaging in the presurgical evaluation of refractory focal epilepsy J Neurol NeurosurgPsychiatry 201283417ndash423

7 Brodbeck V Spinelli L Lascano AM et al Electroencephalographic source imaginga prospective study of 152 operated epileptic patients Brain 20111342887ndash2897

8 Beniczky S Lantz G Rosenzweig I et al Source localization of rhythmic ictal EEGactivity a study of diagnostic accuracy following STARD criteria Epilepsia 2013541743ndash1752

9 Bossuyt PM Cohen JF Gatsonis CA Korevaar DA STARD Group STARD 2015updated reporting guidelines for all diagnostic accuracy studies Ann Transl Med2016485

10 Lantz G Spinelli L Seeck M De Peralta Menendez RG Sottas CC Michel CMPropagation of interictal epileptiform activity can lead to erroneous sourcelocalizations a 128-channel EEG mapping study J Clin Neurophysiol 200320311ndash319

11 Almubarak S Alexopoulos A Von-Podewils F et al The correlation of magneto-encephalography to intracranial EEG in localizing the epileptogenic zone a study ofthe surgical resection outcome Epilepsy Res 20141081581ndash1590

12 Taulu S Kajola M Simola J Suppression of interference and artifacts by the signalspace separation method Brain Topogr 200416269ndash275

13 Bagic AI Knowlton RC Rose DF Ebersole JS American Clinical Magneto-encephalography Society Clinical Practice Guideline 1 recording and analysis ofspontaneous cerebral activity J Clin Neurophysiol 201128348ndash354

14 Birot G Spinelli L Vulliemoz S et al Head model and electrical source imaginga study of 38 epileptic patients Neuroimage Clin 2014577ndash83

15 Beniczky S Aurlien H Brogger JC et al Standardized Computer-Based OrganizedReporting of EEG SCORE Epilepsia 2013541112ndash1124

16 Wagner M Fuchs M Kastner J Evaluation of sLORETA in the presence of noise andmultiple sources Brain Topogr 200416277ndash280

17 Fuchs M Wagner M Wischmann HA et al Improving source reconstructions bycombining bioelectric and biomagnetic data Electroencephalogr Clin Neurophysiol199810793ndash111

18 Fuchs M Kastner J Wagner M Hawes S Ebersole JS A standardized boundaryelement method volume conductor model Clin Neurophysiol 2002113702ndash712

19 SeeckM Koessler L Bast T et al The standardized EEG electrode array of the IFCNClin Neurophysiol 20171282070ndash2077

20 Gaspard N Hirsch LJ LaRoche SM Hahn CD Brandon M Interrater agreement forcritical care EEG terminology Epilepsia 2014551366ndash1373

21 Landis JR Koch GG The measurement of observer agreement for categorical dataBiometrics 197733159ndash174

22 McCullagh P Nelder JA Generalized Linear Models 2nd ed London Chapman ampHall 1989

23 Agresti A Categorical Data Analysis 2nd ed Hoboken NJ JohnWiley amp Sons 200270ndash114

24 Fisher RS Scharfman HE deCurtis M How can we identify ictal and interictalabnormal activity Adv Exp Med Biol 20148133ndash23

25 Lantz G Grave de Peralta R Spinelli L Seeck M Michel CM Epileptic sourcelocalization with high density EEG how many electrodes are needed Clin Neuro-physiol 200311463ndash69

26 Huang MX Song T Hagler DJ Jr et al A novel integrated MEG and EEG analysismethod for dipolar sources Neuroimage 200737731ndash748

27 Aydin U Vorwerk J Kupper P et al Combining EEG andMEG for the reconstructionof epileptic activity using a calibrated realistic volume conductor model PLoS One20149e93154

28 Bast T Oezkan O Rona S et al EEG andMEG source analysis of single and averagedinterictal spikes reveals intrinsic epileptogenicity in focal cortical dysplasia Epilepsia200445621ndash631

e586 Neurology | Volume 92 Number 6 | February 5 2019 NeurologyorgN

DOI 101212WNL0000000000006877201992e576-e586 Published Online before print January 4 2019Neurology Lene Duez Hatice Tankisi Peter Orm Hansen et al

prospective studyElectromagnetic source imaging in presurgical workup of patients with epilepsy A

This information is current as of January 4 2019

ServicesUpdated Information amp

httpnneurologyorgcontent926e576fullincluding high resolution figures can be found at

References httpnneurologyorgcontent926e576fullref-list-1

This article cites 26 articles 1 of which you can access for free at

Subspecialty Collections

httpnneurologyorgcgicollectionfunctional_neuroimagingFunctional neuroimaging

httpnneurologyorgcgicollectionepilepsy_surgery_Epilepsy surgery

httpnneurologyorgcgicollectioneeg_EEG

httpnneurologyorgcgicollectioncortical_localizationCortical localization

httpnneurologyorgcgicollectionclass_ivClass IVfollowing collection(s) This article along with others on similar topics appears in the

Permissions amp Licensing

httpwwwneurologyorgaboutabout_the_journalpermissionsits entirety can be found online atInformation about reproducing this article in parts (figurestables) or in

Reprints

httpnneurologyorgsubscribersadvertiseInformation about ordering reprints can be found online

ISSN 0028-3878 Online ISSN 1526-632XWolters Kluwer Health Inc on behalf of the American Academy of Neurology All rights reserved Print1951 it is now a weekly with 48 issues per year Copyright Copyright copy 2019 The Author(s) Published by

reg is the official journal of the American Academy of Neurology Published continuously sinceNeurology