Magnetoencephalography (MEG) and high-density electroencephalography (HD-EEG) are rarely recorded simultaneously, although they yield confirmatory and complementary information. Here, we illustrate the experimental setup for recording simultaneous MEG and HD-EEG and the methodology for analyzing these data aiming to localize epileptogenic and eloquent brain areas in children with drug-resistant epilepsy.
For children with drug-resistant epilepsy (DRE), seizure freedom relies on the delineation and resection (or ablation/disconnection) of the epileptogenic zone (EZ) while preserving the eloquent brain areas. The development of a reliable and noninvasive localization method that provides clinically useful information for the localization of the EZ is, therefore, crucial to achieving successful surgical outcomes. Electric and magnetic source imaging (ESI and MSI) have been increasingly utilized in the presurgical evaluation of these patients showing promising findings in the delineation of epileptogenic as well as eloquent brain areas. Moreover, the combination of ESI and MSI into a single solution, namely electromagnetic source imaging (EMSI), performed on simultaneous high-density electroencephalography (HD-EEG) and magnetoencephalography (MEG) recordings has shown higher source localization accuracy than either modality alone. Despite these encouraging findings, such techniques are performed in only a few tertiary epilepsy centers, are rarely recorded simultaneously, and are underutilized in pediatric cohorts. This study illustrates the experimental setup for recording simultaneous MEG and HD-EEG data as well as the methodological framework for analyzing these data aiming to localize the irritative zone, the seizure onset zone, and eloquent brain areas in children with DRE. More specifically, the experimental setups are presented for (i) recording and localizing interictal and ictal epileptiform activity during sleep and (ii) recording visual-, motor-, auditory-, and somatosensory-evoked responses and mapping relevant eloquent brain areas (i.e., visual, motor, auditory, and somatosensory) during visuomotor task, as well as auditory and somatosensory stimulations. Detailed steps of the data analysis pipeline are further presented for performing EMSI as well as individual ESI and MSI using equivalent current dipole (ECD) and dynamic statistical parametric mapping (dSPM).
Epilepsy is one of the most common and disabling neurological disorders characterized by recurrent and unprovoked seizures that can be either focal or generalized in nature. Despite the availability of several effective pharmacologic therapies (e.g., antiseizure medications [ASMs]), about 20-30% of these patients are unable to control their seizures and suffer from drug-resistant epilepsy (DRE)1. For these patients, epilepsy surgery is the most effective treatment to eliminate seizures; a successful surgery can be achieved through the complete resection (or ablation/disconnection) of the epileptogenic zone (EZ), defined as the minimal area indispensable for the generation of seizures2. Accurate delineation and resection (or ablation/disconnection) of the EZ while preserving the eloquent cortex are crucial factors in ensuring seizure freedom. To establish surgical candidacy, several noninvasive diagnostic tools are used by a multidisciplinary team to define different cortical areas (i.e., irritative zone, seizure onset zone [SOZ], functional deficit zone, and epileptogenic lesion), which serve as indirect approximators of the EZ3. Extra-operative monitoring with intracranial EEG (iEEG) is required when none of these methods unequivocally identify the EZ. The role of iEEG is to define precisely the EZ by localizing the SOZ (i.e., the brain area where clinical seizures generate) and map eloquent brain areas. Yet, it presents serious limitations due to its invasiveness4,5,6, it offers limited spatial coverage, and it needs a clear presurgical localization hypothesis7. As a result, the actual focus and extent of the SOZ may be missed, leading to unsuccessful surgery. Also, its interpretation requires the recording of multiple stereotyped clinical seizures during several days of hospitalization, which increases the chances of complications (e.g., infection and/or bleeding)5. Hence, there is an unmet need to develop reliable and noninvasive localization methods that can provide clinically useful information and overall improve the presurgical evaluation of children with DRE.
Over the last decades, electric and magnetic source imaging (ESI and MSI) have been increasingly utilized in the presurgical evaluation of patients with DRE for the delineation of epileptogenic as well as functional brain areas. Particularly, ESI and MSI allow the reconstruction of neural sources from noninvasive recordings, such as high-density EEG (HD-EEG) and magnetoencephalography (MEG), to help guide surgical planning or iEEG electrode placement. ESI and MSI can be applied for localizing either interictal epileptiform discharges (IEDs), such as spikes and sharp waves, or ictal (seizure) activity. It may further be used for the localization of different functional brain areas involved in sensory, motor, auditory, and cognitive functions. The reconstruction of electrophysiological events, such as IEDs and seizures, allows the identification of the irritative zone (i.e., the brain area where IEDs originate) and the SOZ, respectively, which are considered a valid surrogate for the EZ localization. The localization of the eloquent cortex (i.e., the brain areas indispensable for defined cortical functions)3 allows instead to map the location and extent of eloquent areas with respect to the planned resection and, therefore, reduce in advance potential functional deficits that may be expected from epilepsy surgery8,9,10,11. Several studies investigated the clinical utility of ESI and/or MSI in the presurgical evaluation of epilepsy, and they showed promising findings in the delineation of the EZ12,13,14,15,16,17,18,19. For example, Mouthaan et al.14 performed an extensive meta-analysis using noninvasive data of 11 prospective and retrospective epilepsy studies and reported that these source localization techniques can overall identify the EZ with high sensitivity (82%) and low specificity (53%). Other studies also showed that MSI and ESI can correctly localize the epileptic focus within the resected area in epileptic patients having a normal magnetic resonance imaging (MRI)19,20,21. These localization results are particularly important for those patients who are ineligible for epilepsy surgery due to inconclusive clinical or imaging findings. In summary, ESI and MSI can contribute significantly to the presurgical mapping of epileptogenic as well as functional brain areas in patients with DRE.
Despite these encouraging findings, such techniques are currently performed in only a few tertiary epilepsy centers on a regular basis and are often underutilized in pediatric populations. Moreover, HD-EEG and MEG are rarely recorded simultaneously, although they provide both confirmatory and complementary information. MEG is sensitive to detect superficial sources with tangential orientation but is blind to radially oriented sources located at the gyri or deeper areas of the brain22,23,24,25,26. Furthermore, MEG provides better spatial resolution (millimeters) compared to EEG16,22,25. Unlike EEG signals, MEG signals are reference-free and are essentially unaffected by different conductivities of brain tissues (i.e., meninges, cerebrospinal fluid, skull, and scalp)25,27 providing undistorted measurements of the magnetic fields produced by the brain. On the other hand, EEG can detect sources of all orientations, but it offers lower spatial resolution than MEG and is more susceptible to artifacts26,28. Due to these complementary sensitivities to source orientation and depth, approximately 30% of the epileptiform activity (e.g., IEDs) can only be recorded on MEG but not on EEG, and vice versa26,29,30,31,32. Contrary to EEG, which allows prolonged recordings, capturing clinical seizures with MEG is challenging due to the restricted recording time that is usually insufficient to record ictal events in most patients. Furthermore, artifacts caused by seizure-related head movements can often interfere with the quality of MEG recordings29,33,34,35. On the other hand, MEG recordings are faster and easier compared to EEG, especially in children since there is no need to attach sensors over the children's head35.
Advances in hardware have made it possible to record simultaneously MEG and HD-EEG data with a high number of sensors (over 550 sensors) covering the whole head. Moreover, modern developments in EEG technologies have minimized the preparation time for HD-EEG to less than a quarter of an hour36. This is particularly important for pediatric populations with challenging behaviors who are unable to stay still for prolonged periods. Furthermore, advancements in software technologies have allowed the combination of ESI and MSI into a single solution, namely electromagnetic source imaging (EMSI), performed on simultaneous HD-EEG and MEG recordings. Several theoretical and empirical studies reported higher source localization accuracy with EMSI than either modality alone13,30,31,37,38,39,40,41. Using different source localization approaches to reconstruct the activity in response to sensory stimuli, Sharon et al.37 found that EMSI had consistently better localization results than either ESI or MSI alone compared to functional MRI (fMRI), which serves as noninvasive benchmark of precise localization accuracy. The authors suggested that this improved localization is due to the increased number of sensors for solving the inverse solution and the different sensitivity patterns of the two imaging modalities37. Similarly, Yoshinaga et al.31 performed dipole analysis on simultaneous EEG and MEG data of patients with intractable localization-related epilepsy and showed that EMSI provided information that would be not obtainable by using only one modality alone and led to successful localization for epilepsy surgery in one of the patients analyzed. In a prospective blinded study, Duez et al.13 showed that EMSI achieved a significantly higher odds ratio (i.e., probability of becoming seizure-free) compared to ESI and MSI, a localization accuracy ≥52%, and a concordance ≥53% and ≥36% with the irritative and SOZ, respectively. A more recent study from our group42 has shown that EMSI provided superior localization estimates and better outcome prediction performance than either ESI or MSI alone, with localization errors from resection and SOZ of ~8 mm and ~15 mm, respectively. Despite these promising findings, there is a lack of studies that provide the methodological framework regarding EMSI in children with DRE.
This study illustrates the experimental setup for performing simultaneous MEG and HD-EEG recordings as well as the methodological framework for analyzing these data aiming to localize the irritative zone, SOZ, and eloquent brain areas in children with DRE. More specifically, the experimental setups are presented for (i) recording and localizing interictal and ictal epileptiform activity during sleep and (ii) recording visual-, motor-, auditory-, and somatosensory-evoked responses and mapping relevant eloquent brain areas (i.e., visual, motor, auditory, and somatosensory) during a visuomotor task, as well as auditory and somatosensory stimulations. Detailed steps of the data analysis pipeline are further presented for performing EMSI as well as individual ESI and MSI using equivalent current dipole (ECD) and dynamic statistical parametric mapping (dSPM).
The experimental procedures applied here have been approved by the North Texas Regional Institutional Review Board (2019-166; Principal Investigator: Christos Papadelis). The following section will describe the experimental protocol for the noninvasive source localization of IEDs, ictal onsets, and event-evoked responses (i.e., visual, motor, auditory, and somatosensory) using simultaneous MEG and HD-EEG recordings followed in our laboratory. The International Federation of Clinical Neurophysiology43 and the American Clinical MEG Society44 have provided "minimum standards" for the routine clinical recording and analysis of spontaneous MEG and EEG data. Procedures for HD-EEG recordings described here apply only to sponge-based EEG electrode systems. The overall preparation process for each subject is about 2-3 h, comprising the actual recordings of ~1.5 h.
1. MEG system's preparation
2. Subject's preparation
3. Data acquisition
NOTE: The acquisition of simultaneous MEG and EEG data is performed in the MEG facility at Cook Children's Medical Center (CCMC). More details about the clinical use of MEG on pediatric children with epilepsy can be found elsewhere8,27,45.
4. Data analysis
Pediatric patients with DRE were recruited from the Epilepsy Clinic at Jane and John Justin Institute for Mind Health, Cook Children's Health Care System (CCHCS). Here, data from three representative patients are presented: (i) a 10-year-old female, (ii) a 13-year-old male, and (iii) a 10-year-old female.
Case 1: A 10-year-old female was admitted with seizures starting at the age of three years. The patient was suffering from daily seizures even after the administration of 8 ASMs. Initial seizures were characterized by eye deviation (unclear side) and behavioral arrest. Later, the patient experienced daily seizures of ~30 s characterized by ictal pouting ("chapeau de gendarme" sign), head deviation to the left, and bilateral tonic arm stiffening (right predominance). Long-term video EEG revealed two clusters of asymmetric tonic seizures with head deviation to the left, followed by her left arm coming up. Three tonic seizures were also observed while sleeping, with frequent runs of generalized fast polyspikes and slow waves with intermittent eye-opening, upward gaze, and left or right arm elevation. These polyspikes and slow-sleep waves were mostly prominent from the left middle temporal lobe. Brain MRI revealed the following multifocal dysplasias: (i) left parietal lobe (postcentral gyrus) focal cortical dysplasia (FCD) with transmantle sign (type II FCD), (ii) right parietooccipital junction FCD, and (iii) left temporal pole FCD. Positron emission tomography (PET) demonstrated hypometabolism in the left parietal lobe, left temporal lobe, and right parietooccipital junction corresponding to the foci of the signal abnormality (i.e., FCD) on the MRI exam. The patient was diagnosed with intractable epilepsy, with stereotyped semiology of chapeau followed by tonic arm stiffing, suggesting possible mesial frontal or insular/temporal onset. Extensive bilateral stereo-EEG (sEEG) exploration was recommended by targeting the frontal lobe, cingulate, insula, and those regions of dysplasia. During iEEG monitoring, the patient had typical seizures with "chapeau de gendarme" followed by tonic elevation/flexion of the right or left upper extremity characterized by diffuse EEG onset, maximal over the bilateral anterior insula. Multifocal IEDs were mostly observed at the right and left anterior temporal lobe and dorsolateral frontal cortex, including the bilateral insula. ESI performed on iEEG recording confirmed the location of SOZ, which was clinically defined bilaterally at the left and right dorsolateral frontal cortex and anterior insula.
As part of the presurgical evaluation, source localization on the simultaneous MEG and HD-EEG data was performed. MEG and HD-EEG recordings indicated frequent IEDs at both frontotemporal regions. Figure 3A shows a representative example of an IED on both MEG and HD-EEG data; topographic field and potential mapping from both modalities indicated a possible underlying source in the right frontotemporal region. ESI indicated a scattered cluster of dipoles covering areas of the right and left frontotemporal and parietal lobes. MSI showed a focal cluster of dipoles in the right frontotemporal lobe, located near the right insula. EMSI indicated focal clusters of dipoles at the bilateral frontotemporal regions, in line with ESI performed on iEEG gold standard, which confirmed the clinical observations (Figure 3C). These dipoles estimated through EMSI showed a mean distance from the iEEG-defined SOZ of 9.81 mm (median: 11.18; std: 2.37).
Case 2: A 13-year-old male with intractable epilepsy was admitted with seizures starting at the age of nine years. Seizures started with an aura followed by leftward head/eye deviation with preserved awareness at times and focal clonus of the head to the left, last for ~30 s, and occurred several times per week. None of the ASMs prescribed achieved seizure control. From the long-term video-EEG, we observed right posterior temporal spikes and frequent spike-wave discharges in the right hemisphere involving the middle temporal, frontotemporal, temporoparietal, and centroparietal cortex. The patient had six electroclinical seizures characterized by a behavioral change, head/eye deviation to the left with left arm extension, and sometimes clonic activity of the left arm, and three seizures with secondary bilateral convulsive activity. The maximal onset was at the right middle temporal lobe with an evolution in the right frontotemporal lobe. Brain MRI revealed an extensive malformation of the cortex in the right cerebral hemisphere (perisylvian predominant) and a mild volume loss in the right cerebral hemisphere with ex vacuo dilation of the right lateral ventricle. The patient was diagnosed with intractable epilepsy with onset in the right hemisphere, favoring the temporal and perisylvian onset in the region of diffuse cortical malformation. Stereo-EEG was performed to delineate the extent of involvement, with electrodes placed in the right temporal, perisylvian, insular, and parietooccipital cortices. Several electroclinical focal onset seizures were captured during the iEEG monitoring with maximal onsets in a wide area of the right frontotemporal lobe. ESI performed on iEEG data localized these seizures in a more focal area comprising both the right temporal (near the right middle temporal gyrus) and perisylvian areas.
As part of the presurgical evaluation, simultaneous MEG and HD-EEG were performed during which the patient experienced two seizures: one while sitting on the wooden chair during the digitization process and one captured during the actual recording with the onset visible on both MEG and HD-EEG (Figure 4A). Topography field and potential maps at the ictal onset indicated that the underlying generator of the seizure onset may be at the right middle temporal lobe, as displayed in Figure 4A. Source localization on the ictal event presented different findings for ESI and MSI: ESI showed dipoles localized toward the right frontotemporal and centroparietal lobes, whereas MSI showed dipoles with high clusterness mostly at the right temporal lobe (Figure 4B), with additional scattered dipoles in the frontotemporal cortex. By combining these solutions, EMSI revealed localization of the ictal onset within the temporal lobe concordant with the ESI on iEEG gold standard (Figure 4B). Particularly, EMSI presented localization results with a mean distance from the SOZ defined by the iEEG monitoring of 12.21 mm (median: 13.62; std: 2.37).
Case 3: A 15-year-old female with localization-related idiopathic epilepsy was admitted with seizures starting at the age of 13 years, but possibly at 8-9 in retrospect, when she was diagnosed with tics due to repetitive, stereotyped neck movements. The patient had brief head tilts to the left that sometimes progressed to focal dyscognitive seizure with hypermotor behaviors (i.e., generalized tonic-clonic seizures), as well as nocturnal convulsive seizures. Several ASMs were administered without achieving complete seizure control. During long-term video-EEG monitoring, the patient had focal electroclinical seizures with secondary generalization with onset at the left posterior temporal lobe, numerous brief focal motor seizures with head tilt to the left, and a subtle electrographic seizure with onset at the left centroparietal cortex. Brain MRI revealed no acute intracranial abnormality and a Chiari I malformation. Positron emission tomography-computed tomography (PET-CT) exam of the head resulted negative. Additional testing, such as ictal single-photon emission CT (SPECT), simultaneous MEG and HD-EEG, cervical spine X-ray, magnetic resonance angiography (MRA) of the head and neck, and eventually sEEG exploration of the left hemisphere, was recommended.
As part of the evaluation, the patient participated in simultaneous MEG and HD-EEG recordings for mapping eloquent brain areas, such as the primary visual, motor, auditory, and somatosensory cortices. Initially, the patient performed a visuomotor task, followed by auditory and somatosensory stimulations. The first cortical response to the visual stimulation occurred at ~70 ms after the stimulus onset for both MEG and HD-EEG (Figure 5A). Figure 5B reports the topography field and potential maps of the cortical locations involved in the visual stimulation for MEG and HD-EEG, respectively. For HD-EEG, a change of polarity in the channels covering the occipital brain areas was observed, whereas a more complex field distribution was found in the same areas for MEG (Figure 5B). Source localization using dSPM revealed a focal cortical activity at this time point within the following brain regions of the Desikan-Killiany atlas: (i) cuneus for MSI; (ii) lateral occipital cortex for ESI; and (iii) cuneus and lateral occipital cortex for EMSI (Figure 5C). Time-frequency analysis on visual cortical responses revealed an event-related synchronization (ERS) in the gamma frequency band for MSI (approximate range: 30-50 Hz), ESI (approximate range: 40-50 Hz), and EMSI (approximate range: 30-50 Hz) (Figure 5D). For the motor-evoked responses, suppression of the mu-rhythm activity was observed over the contralateral M1 during the movement onset (Figure 6A). In Figure 6B, we reported the topography field and potential maps of the brain areas activated during the motor task for MEG and HD-EEG, respectively. MEG field maps indicated clear alterations of magnetic influx and outflux in the contralateral central brain areas, which may indicate an underlying focal generator in the contralateral M1 (Figure 6B). HD-EEG potential maps showed a focal polarity change in the same areas, with electric potentials perpendicular to the magnetic fields (Figure 6B). The peaks of maximal source activation were observed while performing the tapping task at the contralateral precentral gyrus of the Desikan-Killiany atlas for MSI, ESI, and EMSI, respectively, as displayed in Figure 6C. Motor-related cortical responses occurring during the anticipation of the upcoming tapping movement showed ERS in beta and gamma bands for MSI (approximate range: 20-30 Hz) and EMSI (approximate range: 20-40 Hz) and gamma band for ESI (approximate range: 30-50 Hz), referred in the literature as mu rhythm suppression (Figure 6D).55,56 Auditory-evoked fields and potentials in response to auditory stimulation had a maximum positive peak at ~80 ms and ~120 ms after the stimulus onset delivery for MEG and HD-EEG, respectively (Figure 7A). In Figure 7B, we reported the topography field and potential maps of the cortical locations involved in the auditory stimulation for MEG and HD-EEG, respectively. In both MEG and HD-EEG, an obvious polarity change with clearly defined negative and positive poles at the sensors covering the left temporal brain areas was observed; these perpendicular magnetic field and electric potential maps may reveal an underlying focal generator in V1 (Figure 7B). Performing source localization on the averaged auditory-evoked fields and potentials, maximal cortical activation was observed at the transverse temporal gyrus and posterior portion of the superior temporal gyrus of the Desikan-Killiany atlas for MSI, ESI, and EMSI, respectively (Figure 7C). Time-frequency analysis of auditory-evoked responses revealed ERS in the gamma band for MSI (approximate range: 40-60 Hz) and EMSI (approximate range: 35-50 Hz), and beta and gamma frequency bands (approximate range: 25-60 Hz) for ESI (Figure 7D). Finally, we observed the first cortical activity in response to the tactile stimulation at ~60 and ~50 ms after the stimulus onset for MEG and HD-EEG, respectively (Figure 8A). In Figure 8B, we reported the topography field and potential maps of the brain areas activated during the somatosensory stimulation for MEG and HD-EEG, respectively. MEG field maps revealed a clear polarity change with distinct alterations of magnetic flux at sensors covering the contralateral parietal areas, whereas HD-EEG potential maps showed a less obvious change of polarity in the same areas with a stronger positive pole than the negative one. These perpendicular magnetic field and electric potential maps may indicate a focal cortical generator in S1. Using dSPM on the averaged somatosensory-evoked responses, maximal cortical source activity at this time point was observed within the contralateral postcentral gyrus of the Desikan-Killiany atlas for MSI, ESI, and EMSI, respectively (Figure 8C). In response to the tactile stimuli, ERS in beta and gamma frequency bands for MSI (approximate range: 15-40 Hz) and EMSI (approximate range: 20-40 Hz), and gamma frequency band for ESI (approximate range: 30-40 Hz) (Figure 8D) were also observed.
Figure 1: Experimental setup for simultaneous MEG and HD-EEG at CCHCS. (A) HD-EEG (256 channels) and MEG (306 sensors) systems with the gantry of the MEG set to a supine position (90°, horizontal position) for a resting/sleeping state recording using the nonmagnetic MEG-compatible bed. The technician is preparing the subject (a 9-year-old girl) for the recording while ensuring safety and comfort. (B) HD-EEG and MEG systems set for a recording in a seated position using the nonmagnetic MEG-compatible chair. The technician is preparing the subject for the recording while ensuring the correct position of the subject in front of the screen where visual stimuli will be projected during the visuomotor task. Please click here to view a larger version of this figure.
Figure 2: Technical aspects of combining data from simultaneous MEG and HD-EEG recordings using different acquisition systems. (A) Spatial alignment (coregistration) of MEG and HD-EEG sensors into the same coordinate system (defined by subject's head coordinates) for a representative subject (a 9-year-old girl). The head coordinates of the subject are represented by the following fiducial points: nasion (green-colored) and left/right preauricular points (red- and blue-colored, respectively). The 306 MEG sensors (blue-colored) – 102 magnetometers and 204 planar gradiometers – and the head position indicator (HPI) coils (magenta-colored) are displayed; aligned into the same coordinate system, the 256 HD-EEG channels are also displayed (pink-colored). (B) Left panel: Linear drift (i.e., delta, displayed as a black line) of data samples occurring between MEG and HD-EEG systems for a representative subject (a 9-year-old girl). Delta is defined as the absolute value of the difference between the times in which the same trigger is sent to both MEG and EEG systems and continuously increases over time: from low (delta = 0 ms) to high (delta = 197 ms) values. Correction of the linear drift estimated using a polynomial function to be applied to the signals is displayed with a blue dashed line. Corrected drift (delta ~0 ms over time) representing a synchronized time between MEG and EEG systems is displayed with a red dashed line. Right panel: Graphical representation of the time shift (delta = 197 ms) estimated for the last trigger sent to both MEG and EEG systems is displayed. Please click here to view a larger version of this figure.
Figure 3: Interictal epileptiform discharges (IEDs) on MEG and HD-EEG data. (A) Time portion of simultaneous MEG and HD-EEG recording (10 s) from a 10-year-old female (Case 1) with frequent IEDs. A subgroup of the 306 MEG sensors and 256 EEG electrodes has been selected for visualization purposes. Topography field and potential maps at the peak of an IED are displayed as inner panels for MEG and HD-EEG, respectively. (B) Location of MEG and HD-EEG sensors (yellow-colored) coregistered on subject's 3D head and cortical (blue-colored) surfaces. Realistic boundary element method (BEM) head model consisting of three layers [i.e., scalp (grey-colored), outer skull (yellow-colored), and inner skull (pink-colored)] reconstructed from the pre-operative MRI of the subject. (C) Source localization clusterness results performed on IEDs using equivalent current dipole (ECD) are shown on subject's pre-operative MRI for ESI, MSI, EMSI, and ESI on iEEG (gold standard)52. Heat maps of dipole clusterness with a goodness-of-fit >60% are displayed from lower (blue) to higher (red) values. The seizure onset zone defined through ESI performed on iEEG data was regarded as the gold standard (orange and green circles). Please click here to view a larger version of this figure.
Figure 4: Seizure onset on MEG and HD-EEG data. (A) Time portion of simultaneous MEG and HD-EEG recording (10 s) from a 13-year-old male (Case 2) with the seizure onset (red arrow). A subgroup of the 306 MEG sensors and 256 EEG electrodes has been selected for visualization purposes. Topography field and potential maps at the ictal onset are displayed as inner panels for MEG and HD-EEG, respectively. (B) Source localization clusterness results performed at the onset of the ictal event using the equivalent current dipole (ECD) method are shown on the pre-operative MRI of the subject for ESI, MSI, EMSI, and ESI on iEEG (gold standard)52. Heat maps of dipole clusterness with a goodness-of-fit >60% are displayed from lower (blue) to higher (red) values. The seizure onset zone defined through ESI performed on iEEG data was regarded as gold standard (blue circle). Please click here to view a larger version of this figure.
Figure 5: Visual-evoked fields and potentials from MEG and HD-EEG data. (A) Averaged visual-evoked responses of a 15-year-old female for MEG (top panel) and HD-EEG (bottom panel) are displayed for the time interval between -100 ms and 300 ms. (B) Topography field and potential maps of the primary visual cortex are displayed for MEG and HD-EEG, respectively. (C) Source activation maps with maximum amplitudes of cortical activation at brain regions of the Desikan-Killiany atlas (namely, cuneus and lateral occipital cortex) estimated using dynamic statistical parametric mapping (dSPM) method for MSI, ESI, and EMSI, respectively. Heat maps of the source activation (dSPM normalized z-score) are displayed. (D) Time-frequency maps obtained using Morlet wavelet time-frequency decomposition on the visual-evoked responses at the primary visual cortex are displayed for the -100 ms to 300 ms time window. Heat maps of the time-frequency power, expressed in percentages based on the deviation of the normalized data from the mean over the baseline [-200; 0] ms, are displayed. Please click here to view a larger version of this figure.
Figure 6: Motor-evoked fields and potentials from MEG and HD-EEG data. (A) Averaged motor-evoked responses of a 15-year-old female for MEG (top panel) and HD-EEG (bottom panel) are displayed for the left index-tapping task in the time-interval between -100 and 300 ms. The electromyography (EMG) signal (middle panel) with the movement onset (purple arrow) is displayed for the time-interval between -100 ms and 300 ms; the signal is filtered in the frequency band 30-300 Hz (Notch filter: 60 Hz). (B) Topography field and potential maps of the primary motor cortex are displayed for MEG and HD-EEG, respectively. (C) Source activation maps with maximum amplitudes of cortical activation at the contralateral precentral gyrus of the Desikan-Killiany atlas estimated using dynamic statistical parametric mapping (dSPM) method for MSI, ESI, and EMSI, respectively. Heat maps of the source activation (dSPM normalized z-score) are displayed, together with the central sulcus (black line). (D) Time-frequency maps obtained using Morlet wavelet time-frequency decomposition on the motor-evoked responses at the primary motor cortex for the -300 ms to 500 ms time window. Heat maps of the time-frequency power, expressed in percentages based on the deviation of the normalized data from the mean over the baseline [-1500; -1000] ms, are displayed. Please click here to view a larger version of this figure.
Figure 7: Auditory-evoked fields and potentials from MEG and HD-EEG data. (A) Averaged auditory-evoked responses of a 15-year-old female for MEG (top panel) and HD-EEG (bottom panel) are displayed for the time interval between -100 ms and 300 ms. (B) Topography field and potential maps of the primary auditory cortex are displayed for the MEG and HD-EEG, respectively. (C) Source activation maps with maximum amplitudes of cortical activation at the transverse temporal gyrus and posterior portion of the superior temporal gyrus of the Desikan-Killiany atlas estimated using dynamic statistical parametric mapping (dSPM) method for MSI, ESI, and EMSI, respectively. Heat maps of the source activation (dSPM normalized z-score) are displayed. (D) Time-frequency maps obtained using Morlet wavelet time-frequency decomposition on the auditory-evoked responses at the primary auditory cortex for the -100 to 300 ms time window. Heat maps of the time-frequency power, expressed in percentages based on the deviation of the normalized data from the mean over the baseline [-500; 0] ms, are displayed. Please click here to view a larger version of this figure.
Figure 8: Somatosensory-evoked fields and potentials from MEG and HD-EEG data. (A) Averaged somatosensory-evoked responses of a 15-year-old female for MEG (top panel) and HD-EEG (bottom panel) are displayed for the left digits' stimulation in the time interval between -100 and 300 ms. (B) Topography field and potential maps of the primary somatosensory cortex are displayed for the MEG and HD-EEG, respectively. (C) Source activation maps with maximum amplitudes of cortical activation at the contralateral postcentral gyrus of the Desikan-Killiany atlas estimated using dynamic statistical parametric mapping (dSPM) method for MSI, ESI, and EMSI, respectively. Heat maps of the source activation (dSPM normalized z-score) are displayed, together with the central sulcus (black line). (D) Time-frequency maps obtained using Morlet wavelet time-frequency decomposition on the somatosensory-evoked responses at the primary somatosensory cortex for the -100 ms to 300 ms time window. Heat maps of the time-frequency power, expressed in percentages based on the deviation of the normalized data from the mean over the baseline [-100; 0] ms, are displayed. Please click here to view a larger version of this figure.
In this study, we illustrate the experimental setup to record simultaneous MEG and HD-EEG in children with DRE while resting/sleeping, performing a task, or receiving stimuli, and propose a methodological framework for localizing the irritative zone, SOZ, and eloquent brain areas using EMSI, as well as individual MSI and ESI. We further provide technical recommendations for merging MEG and HD-EEG data from different commercially available products that present unique features. We present data from three cases to enhance the clinical utility of EMSI in the localization of epileptogenic and eloquent brain areas. The findings here indicate that EMSI results outperform those obtained by either modality alone, most likely due to the additive value of the complementary properties of MEG and EEG signals in the combined solution and possibly due to the increased number of sensors used for recording the data (>550 sensors). Particularly, EMSI noninvasively localized the irritative and SOZs with concordant findings as ESI on iEEG gold standard, which confirmed the clinical observations.
The proposed methodology includes the following critical steps: (i) high-quality acquisition of simultaneous MEG and HD-EEG (i.e., high SNR) recordings with high spatial sampling of sensors (>550 sensors) covering the entire brain of interictal and ictal activities, as well as visual-, motor-, auditory-, and somatosensory-evoked fields and potentials, from children with DRE (steps 3.1-3.2); (ii) temporal synchronization and spatial co-registration of MEG and HD-EEG signals recorded with different acquisition systems (step 3.12); (iii) careful preprocessing and selection of data portions containing interictal activity (steps 4.1.1-4.1.7), ictal onset activity (steps 4.2.1-4.2.7), and event-related responses (steps 4.3.1-4.3.6), respectively; and (iv) accurate source localization of the irritative zone, SOZ, and eloquent brain areas of interest using reliable source localization methods (e.g., ECDs with clustering and dSPM) (steps 4.1.8-4.1.9, 4.2.8-4.2.9, and 4.3.7-4.3.9, respectively).
The most critical step when performing simultaneous MEG and HD-EEG recordings is to spatially (alignment between coordinate spaces) and temporally (correction of the linear clock drift) synchronize the data recorded by the two acquisition systems. Such synchronization is crucial to ensure the correct identification of interictal, ictal, and visual/motor/auditory/tactile events that simultaneously occur in the MEG and HD-EEG signals. Errors in the timepoint selection of these events may affect the source localization results and identify areas of the brain that are not necessarily involved in the generation of these events.
MEG systems often offer compatible 32-, 64-, and 128-channel EEG systems incorporated into the product for performing simultaneous MEG and EEG measurements. In these cases, there is no need to temporally synchronize the data by sending common trigger signals. Similarly, most EEG systems are nowadays compatible with all MEG systems. Despite these advances in hardware, only few epilepsy centers perform simultaneous MEG and HD-EEG recordings as part of the presurgical evaluation. Here, we took advantage of such integrability and combined the 306-channel MEG and 256-channel EEG systems to simultaneously record the brain activity with >550 sensors covering the subject's head. So far, few software for advanced analysis of MEG, HD-EEG, and iEEG data (e.g., Brainstorm, CURRY, EEGLab, FieldTrip, MNE, or NUTMEG) are available. Future studies are therefore necessary to validate the proposed methodology with new neuroimaging analysis software. Finally, the combination of MSI and ESI into a unique solution (EMSI) increased the computational complexity of the data analysis.
The described method presents a few limitations that should be addressed in future studies. We manually selected IEDs occurring on both MEG and HD-EEG data of two representative patients while ignoring interictal spikes that occurred in only one of the two signals (either MEG or EEG). Manual selection of spikes can be a time-consuming and subjective approach that can be simplified using automated approaches for detecting IEDs developed during the last decades57,58,59. However, visual inspection is always recommended for careful analysis and refined detection of each IED. Furthermore, we used the SOZ as an approximator of the EZ. Yet, the SOZ does not always predict surgical outcomes60,61,62,63. Future studies can, therefore, use the surgical outcome as ground truth for more precise delineation of the EZ13,14,15,16,17,19,20. Although seizures can be successfully captured using simultaneous MEG and EEG and localized using appropriate source localization techniques44,64, it is relatively rare to record such ictal events in clinical practice, especially from outpatients on ASMs. This is mostly due to the limited duration of MEG recordings and the excessive body movements occurring during seizures (e.g., the patient's head slipped out the dewar), which may cause biological artifacts that can severely affect source localization findings. In a recent review, Stefan et al. reported the occurrence of seizures during MEG recordings in 7%-24% of patients, with an average recording time of 30 minutes up to 5.7 h across different studies65. At CCMC, 18 out of 89 (20.2%) patients had ictal events captured during simultaneous MEG and HD-EEG recordings performed within the past ~2 years. However, only 8 out of the 18 (44.4%) patients were successfully analyzed. In cases where interictal MEG recordings show normal or inconclusive findings, ictal MEG or HD-EEG may be used to localize the EZ with high precision. Yet, technical and logistical requirements for these recordings should be addressed. In addition, the representative data for the eloquent cortex localization via EMSI were not compared with any gold standards for localization of these functional brain areas, such as noninvasive fMRI or intraoperative electrocortical stimulation. Further investigations may, therefore, integrate EMSI and fMRI towards a multimodal noninvasive imaging tool to improve the localization accuracy of these eloquent brain areas in children with DRE. This work may also be extended to localize other functional brain areas, such as the language-eloquent regions. Localization of language functions is of critical importance during the presurgical evaluation of patients with DRE to determine their surgical candidacy, plan the extent of surgical resection, and prevent permanent postoperative functional deficits66. Several noninvasive studies have shown that language mapping using MEG can provide concordant results, similar to the invasive Wada test, which is often regarded as the gold standard for identifying the dominant language hemisphere67,68,69,70. A recent study has proposed a multimodal approach in which the combination of different techniques (i.e., cortical stimulation mapping, high-gamma electrocorticography, fMRI, and transcranial magnetic stimulation) can provide mutual, confirmatory, and complementary information for presurgical language mapping71. Despite these advantages, mapping language areas is still challenging in pediatric patients who have cognitive, intellectual, and language barriers due to their age. Thus, more age-specific tasks and child-friendly setups should be developed in the near future. In this work, we analyzed MEG and HD-EEG data using software that is not certified for clinical purposes. Although these tools have been proven to be valuable and effective, they carry liability issues that should be considered when reporting presurgical evaluation findings for clinical use. Here, we describe procedures for HD-EEG recordings using only sponge-based EEG electrode systems. Alternative systems using gel-based EEG electrodes are widely used in both clinical and research settings. Although they provide higher SNR EEG recordings, they required longer preparation time (~40-60 min) and thus are less suitable for pediatric use. Alternatively, several laboratories use low-density gel-based EEG systems during the MEG recordings, which are advantageous in terms of preparation time (compared to HD-EEG systems), but they offer significantly lower spatial resolution due to the reduced number of electrodes covering the entire scalp12,16,72,73.
At present, the localization of the epileptogenic brain areas in patients with epilepsy is still mainly achieved with iEEG monitoring. Moreover, the methodology for the precise localization of eloquent brain areas is poorly defined, and the experiment setups currently used in MEG laboratories are inappropriate for pediatric patients, while the use of HD-EEG for this purpose is very limited. Accurate localization of these areas may facilitate the presurgical evaluation and augment the surgical planning for either resection or iEEG electrode placement. So far, several studies investigated the contribution of either ESI or MSI in the presurgical evaluation of patients with DRE and focal epilepsies for the identification of the EZ12,13,14,15,16,17,18,19 and eloquent areas of the somatosensory cortex41, respectively. Few studies have shown better source localization results and outcome prediction performance using EMSI compared to either MSI or ESI alone13,31,42. Despite these findings, recording of MEG and EEG is rarely performed simultaneously, and MSI and ESI are implemented in only a few epilepsy centers worldwide. To our best knowledge, this is the first study that provides suggestions for collecting and analyzing simultaneous MEG and HD-EEG data, as well as performing EMSI in pediatric epilepsy for the noninvasive identification of the irritative zone, SOZ, and eloquent brain areas, namely primary visual, motor, auditory, and somatosensory cortices.
Here, we performed EMSI on interictal spikes and ictal events detected on simultaneous noninvasive data from two patients with DRE (Cases 1 and 2) and achieved a source localization error of ~9 mm and ~12 mm from the SOZ, respectively, in line with previous studies42. Impressively, such a method achieved a localization accuracy comparable to the intracranial findings (i.e., ESI on iEEG data), with clustered dipoles localized in the brain area pinpointed as epileptogenic by the clinical observations (Figure 3C and Figure 4B). Using noninvasive data from a third representative patient with DRE (Case 3), we also performed EMSI on visual-, motor-, auditory-, and somatosensory-evoked activities and found prominent source activation patterns in the corresponding eloquent brain areas (i.e., visual, motor, auditory, and somatosensory cortices) (Figure 5C, Figure 6C, Figure 7C, and Figure 8C).
Our results were derived from the fusion of complementary information captured from MEG and EEG modalities that may improve localization accuracy. EEG is well known to reflect all intracranial currents, whereas MEG is mostly sensitive to tangential sources and blind to deep brain sources29,74. As shown in this study, combining MEG and EEG can, therefore, overcome the limitations of each modality, provide superior localization results, and identify epileptogenic and eloquent brain areas that either ESI or MSI may have missed if used alone. Moreover, we present an alternative noninvasive approach for mapping eloquent brain areas using EMSI in those patients who did not undergo fMRI during their presurgical evaluation.
The localization of epileptogenic and eloquent brain areas using noninvasive techniques, such as simultaneous MEG and EEG, is an essential step during the presurgical evaluation of children with DRE for complete removal or disconnection of the EZ while preserving eloquent cortical areas. The proposed methodology offers a detailed description of the acquisition and analysis of simultaneous MEG and EEG data that supports its application not only in the presurgical epilepsy evaluation but also in cognitive neurosciences for exploring physiological functions of the healthy brain in both typically developing children as well as healthy adults, as well as morphological and functional brain changes associated with epilepsy or other neurological disorders. Future studies investigating epileptogenic brain networks may also assess whether network hubs (i.e., highly connected brain regions) estimated noninvasively using EMSI on simultaneous MEG and HD-EEG data can more accurately localize the EZ in children with DRE than those estimated using MSI and/or ESI alone75,76,77. Furthermore, the noninvasive mapping of spatiotemporal propagations of spikes and ripples (i.e., high-frequency oscillations, >80 Hz), estimated through EMSI, can help to better understand the pathophysiological mechanisms of propagating epileptiform activity and noninvasively assess the onset generator of these propagations that is a precise biomarker of the EZ78,79. The presented protocol may help to further investigate the complementarity of MEG and EEG systems by examining the sensitivity of MEG and EEG sensor arrays to sources of different orientations. Such analysis may provide insights into the electrophysiological properties of the brain while performing simultaneous MEG and HD-EEG.
The authors have nothing to disclose.
This work was supported by the National Institute of Neurological Disorders and Stroke (R01NS104116; R01NS134944; Principal Investigator: Christos Papadelis).
AIRSTIM unit | SD Instruments | N/A | The SDI AIRSTIM system is an alternative unconditioned stimulus to shock |
Baby Shampoo | Johnson's | N/A | Baby Shampoo is as gentle to the eyes as pure water and is specially designed to gently cleanse baby’s delicate hair and scalp. |
Control III disinfectant cleaning solution | Maril Products, Inc. | http://www.controlthree.com/ | Disinfectant and germicide solution formulated for hospitals |
Elekta Neuromag | TRIUX | NM24132A | Comprehensive bioelectromagnetic measurement system characterized by 306-channel neuromagnetometer for functional brain studies |
FASTRAK | Polhemus technology | NS-7806 | Using A/C electromagnetic technology, FASTRAK delivers accurate position and orientation data, with virtually no latency. With a single magnetic source, FASTRAK delivers data for up to four sensors. The source emits an electromagnetic field, sensors within the field of range are tracked in full 6DOF (6 Degrees-Of-Freedom). Setup is simple and intuitive, with no user calibration required. |
Genuine Grass Reusable Cup EEG Electrodes | Natus Medical, Inc. | N/A | Each Genuine Grass EEG Electrode undergoes rigorous mechanical and electrical testing to assure long life for unsurpassed recording clarity and dependability. |
Geodesic Sensor Net | Electrical Geodesics, Inc. | S-MAN-200-GSNR-001 | 32 to 256 electrodes to place on the human head to aquire dense-array electroencephalography data |
GeoScan Sensor Digitization System | Electrical Geodesics, Inc. | 8100550-03 | Handheld Scanner and Software for 3D electrode position registration |
Natus Xltek NeuroWorks | Natus Medical, Inc. | https://natus.com/ | The Natus NeuroWorks platform simplifies the process of collecting, monitoring and managing data for routine EEG testing, ambulatory EEG, long-term monitoring, ICU monitoring, and research studies. |
Natus NeuroWorks EEG Software | Natus Medical, Inc. | https://natus.com/neuro/neuroworks-eeg-software/ | NeuroWorks EEG software simplifies the process of collecting, monitoring, trending and managing EEG testing data, allowing care providers to save time and focus on delivering the best care. |
ROSA ONE Brain | Zimmer Biomet | https://www.zimmerbiomet.com/en/products-and-solutions/zb-edge/robotics/rosa-brain.html | ROSA ONE Brain is a robotic solution to assist surgeons in planning and performing complex neurosurgical procedures through a small drill hole in the skull. |
Ten20 Conductive Paste | Weaver and company | N/A | Ten20 contains the right balance of adhesiveness and conductivity, enabling the electrodes to remain in place while allowing the transmittance of electrical signals. |