Magnetoencephalography and magnetic source imaging in epilepsy

Deep learning based source imaging provides strong sublobar localization of epileptogenic zone from MEG interictal spikes

Electromagnetic source imaging (ESI) offers unique capability of imaging brain dynamics for studying brain functions and aiding the clinical management of brain disorders. Challenges exist in ESI due to the ill-posedness of the inverse problem and thus the need of modeling the underlying brain dynamics for regularizations. Advances in generative models provide opportunities for more accurate and realistic source modeling that could offer an alternative approach to ESI for modeling the underlying brain dynamics beyond equivalent physical source models. However, it is not straightforward to explicitly formulate the knowledge arising from these generative models within the conventional ESI framework. Here we investigate a novel source imaging framework based on mesoscale neuronal modeling and deep learning (DL) that can learn the sensor-source mapping relationship directly from MEG data for ESI. Two DL-based ESI models were trained based on data generated by neural mass models and either generic or personalized head models. The robustness of the two DL models was evaluated by systematic computer simulations and clinical validation in a cohort of 29 drug-resistant focal epilepsy patients who underwent intracranial EEG (iEEG) evaluation or surgical resection. Results estimated from pre-operative MEG interictal spikes were quantified using the overlap with resection regions and the distance to the seizure-onset zone (SOZ) defined by iEEG recordings. The DL-based ESI provided robust results when no personalized head geometry is considered, reaching a spatial dispersion of 21.90 ± 19.03 mm, sublobar concordance of 83 %, and sublobar sensitivity and specificity of 66 and 97 % respectively. When using personalized head geometry derived from individual patients' MRI in the training data, personalized DL-based ESI model can further improve the performance and reached a spatial dispersion of 8.19 ± 8.14 mm, sublobar concordance of 93 %, and sublobar sensitivity and specificity of 77 and 99 % respectively. When compared to the SOZ, the localization error of the personalized approach is 15.78 ± 5.54 mm, outperforming the conventional benchmarks. This work demonstrates that combining generative models and deep learning enables an accurate and robust imaging of epileptogenic zone from MEG recordings with strong sublobar precision, suggesting its added value to enhancing MEG source localization and imaging, and to epilepsy source localization and other clinical applications.

Addressing a Deep Problem With Magnetoencephalography

[Box: see text].

Applications of Resting State Functional MR Imaging to Traumatic Brain Injury

Traumatic brain injury (TBI) is an important public health issue. TBI includes a broad spectrum of injury severities and abnormalities. Functional MR imaging (fMR imaging), both resting state (rs) and task, has been used often in research to study the effects of TBI. Although rs-fMR imaging is not currently applicable in clinical diagnosis of TBI, computer-aided tools are making this a possibility for the future. Specifically, graph theory is being used to study the change in networks after TBI. Machine learning methods allow researchers to build models capable of predicting injury severity and recovery trajectories.

Current use of imaging and electromagnetic source localization procedures in epilepsy surgery centers across Europe

OBJECTIVE

In 2014 the European Union-funded E-PILEPSY project was launched to improve awareness of, and accessibility to, epilepsy surgery across Europe. We aimed to investigate the current use of neuroimaging, electromagnetic source localization, and imaging postprocessing procedures in participating centers.

METHODS

A survey on the clinical use of imaging, electromagnetic source localization, and postprocessing methods in epilepsy surgery candidates was distributed among the 25 centers of the consortium. A descriptive analysis was performed, and results were compared to existing guidelines and recommendations.

RESULTS

Response rate was 96%. Standard epilepsy magnetic resonance imaging (MRI) protocols are acquired at 3 Tesla by 15 centers and at 1.5 Tesla by 9 centers. Three centers perform 3T MRI only if indicated. Twenty-six different MRI sequences were reported. Six centers follow all guideline-recommended MRI sequences with the proposed slice orientation and slice thickness or voxel size. Additional sequences are used by 22 centers. MRI postprocessing methods are used in 16 centers. Interictal positron emission tomography (PET) is available in 22 centers; all using 18F-fluorodeoxyglucose (FDG). Seventeen centers perform PET postprocessing. Single-photon emission computed tomography (SPECT) is used by 19 centers, of which 15 perform postprocessing. Four centers perform neither PET nor SPECT in children. Seven centers apply magnetoencephalography (MEG) source localization, and nine apply electroencephalography (EEG) source localization. Fourteen combinations of inverse methods and volume conduction models are used.

SIGNIFICANCE

We report a large variation in the presurgical diagnostic workup among epilepsy surgery centers across Europe. This diversity underscores the need for high-quality systematic reviews, evidence-based recommendations, and harmonization of available diagnostic presurgical methods.

MEG-EEG Information Fusion and Electromagnetic Source Imaging: From Theory to Clinical Application in Epilepsy

The purpose of this study is to develop and quantitatively assess whether fusion of EEG and MEG (MEEG) data within the maximum entropy on the mean (MEM) framework increases the spatial accuracy of source localization, by yielding better recovery of the spatial extent and propagation pathway of the underlying generators of inter-ictal epileptic discharges (IEDs). The key element in this study is the integration of the complementary information from EEG and MEG data within the MEM framework. MEEG was compared with EEG and MEG when localizing single transient IEDs. The fusion approach was evaluated using realistic simulation models involving one or two spatially extended sources mimicking propagation patterns of IEDs. We also assessed the impact of the number of EEG electrodes required for an efficient EEG–MEG fusion. MEM was compared with minimum norm estimate, dynamic statistical parametric mapping, and standardized low-resolution electromagnetic tomography. The fusion approach was finally assessed on real epileptic data recorded from two patients showing IEDs simultaneously in EEG and MEG. Overall the localization of MEEG data using MEM provided better recovery of the source spatial extent, more sensitivity to the source depth and more accurate detection of the onset and propagation of IEDs than EEG or MEG alone. MEM was more accurate than the other methods. MEEG proved more robust than EEG and MEG for single IED localization in low signal-to-noise ratio conditions. We also showed that only few EEG electrodes are required to bring additional relevant information to MEG during MEM fusion.

Epileptogenic zone localization using magnetoencephalography predicts seizure freedom in epilepsy surgery

OBJECTIVE

The efficacy of epilepsy surgery depends critically upon successful localization of the epileptogenic zone. Magnetoencephalography (MEG) enables noninvasive detection of interictal spike activity in epilepsy, which can then be localized in three dimensions using magnetic source imaging (MSI) techniques. However, the clinical value of MEG in the presurgical epilepsy evaluation is not fully understood, as studies to date are limited by either a lack of long-term seizure outcomes or small sample size.

METHODS

We performed a retrospective cohort study of patients with focal epilepsy who received MEG for interictal spike mapping followed by surgical resection at our institution.

RESULTS

We studied 132 surgical patients, with mean postoperative follow-up of 3.6 years (minimum 1 year). Dipole source modeling was successful in 103 patients (78%), whereas no interictal spikes were seen in others. Among patients with successful dipole modeling, MEG findings were concordant with and specific to the following: (1) the region of resection in 66% of patients, (2) invasive electrocorticography (ECoG) findings in 67% of individuals, and (3) the magnetic resonance imaging (MRI) abnormality in 74% of cases. MEG showed discordant lateralization in ~5% of cases. After surgery, 70% of all patients achieved seizure freedom (Engel class I outcome). Whereas 85% of patients with concordant and specific MEG findings became seizure-free, this outcome was achieved by only 37% of individuals with MEG findings that were nonspecific to or discordant with the region of resection (χ(2) = 26.4, p < 0.001). MEG reliability was comparable in patients with or without localized scalp electroencephalography (EEG), and overall, localizing MEG findings predicted seizure freedom with an odds ratio of 5.11 (95% confidence interval [CI] 2.23-11.8).

SIGNIFICANCE

MEG is a valuable tool for noninvasive interictal spike mapping in epilepsy surgery, including patients with nonlocalized findings receiving long-term EEG monitoring, and localization of the epileptogenic zone using MEG is associated with improved seizure outcomes.

Neural oscillation, network, eloquent cortex and epileptogenic zone revealed by magnetoencephalography and awake craniotomy

BACKGROUND

Magnetoencephalography (MEG) is a method of functional neuroimaging. The concomitant use of MEG and electrocorticography has been found to be useful in elucidating neural oscillation and network, and to localize epileptogenic zone and functional cortex. We describe our early experience using MEG in neurosurgical patients, emphasizing on its impact on patient management as well as the enrichment of our knowledge in neurosciences.

MATERIALS AND METHODS

A total of 10 subjects were included; five patients had intraaxial tumors, one with an extraaxial tumor and brain compression, two with arteriovenous malformations, one with cerebral peduncle hemorrhage and one with sensorimotor cortical dysplasia. All patients underwent evoked and spontaneous MEG recordings. MEG data was processed at band-pass filtering frequency of between 0.1 and 300 Hz with a sampling rate of 1 kHz. MEG source localization was performed using either overdetermined equivalent current dipoles or underdetermined inversed solution. Neuromag collection of events software was used to study brain network and epileptogenic zone. The studied data were analyzed for neural oscillation in three patients; brain network and clinical manifestation in five patients; and for the location of epileptogenic zone and eloquent cortex in two patients.

RESULTS

We elucidated neural oscillation in three patients. One demonstrated oscillatory phenomenon on stimulation of the motor-cortex during awake surgery, and two had improvement in neural oscillatory parameters after surgery. Brain networks corresponding to clinico-anatomical relationships were depicted in five patients, and two networks were illustrated here. Finally, we demonstrated epilepsy cases in which MEG data was found to be useful in localizing the epileptogenic zones and functional cortices.

CONCLUSION

The application of MEG while enhancing our knowledge in neurosciences also has a useful role in epilepsy and awake surgery.

Neuromagnetic coherence of epileptic activity: an MEG study

PURPOSE

This study was undertaken to test the hypothesis that patients with epilepsy have abnormal imaginary coherence compared with control subjects.

METHODS

Thirty patients with seizures underwent magnetoencephalography (MEG) recording using a whole cortex MEG system. Conventional equivalent current dipoles (ECDs) and synthetic aperture magnetometry (SAM) were used to analyze MEG data. Neural synchronization was studied using imaginary coherence to analyze resting-state MEG data. The ECDs, SAM, and MEG results were then compared with intra/extra-operative EEG.

RESULTS

Abnormal imaginary coherence was identified in all patients (30/30, 100%). The locations of abnormal imaginary coherence were in agreement with the ECDs locations of spikes in 23 patients (23/30, 76.7%). The ECD locations in 5 patients were scattered or located bilaterally. The locations of abnormal imaginary coherence were in agreement with SAM locations in 26 patients (26/30, 86.7%). One case of imaginary coherence was located in two lobes. The ECDs fit locations were in agreement with SAM locations in 21 patients (21/30, 70.0%). The locations of abnormal imaginary coherence, ECDs, and SAM were in agreement with intra/extra-operative EEG in 23 patients (23/30, 76.7%), 17 patients (17/30, 56.7%), and 20 patients (20/30, 66.7%), respectively. The results of ECDs location, SAM location, imaginary coherence, and intracranial EEG (iEEG) were consistent in 15 patients (15/30, 50%).

CONCLUSIONS

The results show that patients with epilepsy have abnormal imaginary coherence, and suggest that the location and coherence of epileptic activity could be quantitatively identified and analyzed using neuromagnetic signals.

Magnetoencephalography in pediatric epilepsy

Magnetoencephalography (MEG) records the magnetic field generated by electrical activity of cortical neurons. The signal is not distorted or attenuated, and it is contactless recording that can be performed comfortably even for longer than an hour. It has excellent and decent temporal resolution, especially when it is combined with the patient's own brain magnetic resonance imaging (magnetic source imaging). Data of MEG and electroencephalography are not mutually exclusive and it is recorded simultaneously and interpreted together. MEG has been shown to be useful in detecting the irritative zone in both lesional and nonlesional epilepsy surgery. It has provided valuable and additive information regarding the lesion that should be resected in epilepsy surgery. Better outcomes in epilepsy surgery were related to the localization of the irritative zone with MEG. The value of MEG in epilepsy surgery is recruiting more patients to epilepsy surgery and providing critical information for surgical planning. MEG cortical mapping is helpful in younger pediatric patients, especially when the epileptogenic zone is close to the eloquent cortex. MEG is also used in both basic and clinical research of epilepsy other than surgery. MEG is a valuable diagnostic modality for diagnosis and treatment, as well as research in epilepsy.

Hypothalamic hamartomas. Part 1. Clinical, neuroimaging, and neurophysiological characteristics

Hypothalamic hamartomas are uncommon but well-recognized developmental malformations that are classically associated with gelastic seizures and other refractory seizure types. The clinical course is often progressive and, in addition to the catastrophic epileptic syndrome, patients commonly exhibit debilitating cognitive, behavioral, and psychiatric disturbances. Over the past decade, investigators have gained considerable knowledge into the pathobiological and neurophysiological properties of these rare lesions. In this review, the authors examine the causes and molecular biology of hypothalamic hamartomas as well as the principal clinical features, neuroimaging findings, and electrophysiological characteristics. The diverse surgical modalities and strategies used to manage these difficult lesions are outlined in the second article of this 2-part review.

SQUID-detected ultra-low field MRI

MRI remains the premier method for non-invasive imaging of soft-tissue. Since the first demonstration of ULF MRI the trend has been towards ever higher magnetic fields. This is because the signal, and efficiency of Faraday detectors, increases with ever higher magnetic fields and corresponding Larmor frequencies. Nevertheless, there are many compelling reasons to continue to explore MRI at much weaker magnetic fields, the so-called ultra-low field or (ULF) regime. In the past decade many excellent proof-of-concept demonstrations of ULF MRI have been made. These include combined MRI and magnetoencephalography, imaging in the presence of metal, unique tissue contrast, and implementation in situations where a high magnetic field is simply impractical. These demonstrations have routinely used pulsed pre-polarization (at magnetic fields from ~10 to 100 mT) followed by read-out in a much weaker (1-100 μT) magnetic fields using the ultra-sensitive Superconducting Quantum Interference Device (SQUID) sensor. Even with pre-polarization and SQUID detection, ULF MRI suffers from many challenges associated with lower magnetization (i.e. signal) and inherently long acquisition times compared to conventional >1 T MRI. These are fundamental limitations imposed by the low measurement and gradient fields used. In this review article we discuss some of the techniques, potential applications, and inherent challenges of ULF MRI.

SQUID-detected ultra-low field MRI

MRI remains the premier method for non-invasive imaging of soft-tissue. Since the first demonstration of ULF MRI the trend has been towards ever higher magnetic fields. This is because the signal, and efficiency of Faraday detectors, increases with ever higher magnetic fields and corresponding Larmor frequencies. Nevertheless, there are many compelling reasons to continue to explore MRI at much weaker magnetic fields, the so-called ultra-low field or (ULF) regime. In the past decade many excellent proof-of-concept demonstrations of ULF MRI have been made. These include combined MRI and magnetoencephalography, imaging in the presence of metal, unique tissue contrast, and implementation in situations where a high magnetic field is simply impractical. These demonstrations have routinely used pulsed pre-polarization (at magnetic fields from ∼10 to 100mT) followed by read-out in a much weaker (1-100μT) magnetic fields using the ultra-sensitive Superconducting Quantum Interference Device (SQUID) sensor. Even with pre-polarization and SQUID detection, ULF MRI suffers from many challenges associated with lower magnetization (i.e. signal) and inherently long acquisition times compared to conventional >1T MRI. These are fundamental limitations imposed by the low measurement and gradient fields used. In this review article we discuss some of the techniques, potential applications, and inherent challenges of ULF MRI.

Tangential and radial epileptic spike activity: different sensitivity in EEG and MEG

OBJECTIVE

Observations in epileptic patients show that interictal spikes are sometimes only visible in electroencephalography (EEG) and sometimes only in magnetoencephalography (MEG). This observation cannot readily be explained by the theoretical sensitivities of EEG and MEG based on analytical models. In this context, we aimed to study the directional sensitivity of radial and tangential spike activity in numerical simulations using realistic head models.

METHODS

We calculated the signal-to-noise ratio (SNR) of simulated spikes at varying orientations and with varying background activity in 12 brain regions in 4 volunteers. Different levels of background activity were modeled by adjusting the amplitudes of several thousand dipoles distributed in the cortex.

RESULTS

For a fixed realistic background activity, we found a higher SNR for MEG spikes for spike orientations that deviated not > 30° from the tangential direction. In contrast, we found a higher SNR for EEG spikes that deviated not > 45° from the radial direction. When the radial background activity was selectively increased, the sensitivity of EEG for radially oriented spikes decreased; when the tangential background activity selectively increased, the sensitivity of MEG for tangentially oriented spikes was decreased.

CONCLUSIONS

Our simulations provide a possible explanation for the clinically observed differences in epileptic spike detection between EEG and MEG. Epileptic spike detection can be improved by analyzing a combination of EEG and MEG data.

Magnetoencephalography in pediatric lesional epilepsy surgery

This study was performed to assess the usefulness of magnetoencephalography (MEG) as a presurgical evaluation modality in Korean pediatric patients with lesional localization-related epilepsy. The medical records and MEG findings of 13 pediatric patients (6 boys and 7 girls) with localization-related epilepsy, who underwent epilepsy surgery at Seoul National University Children's Hospital, were retrospectively reviewed. The hemispheric concordance rate was 100% (13/13 patients). The lobar or regional concordance rate was 77% (10/13 patients). In most cases, the MEG spike sources were clustered in the proximity of the lesion, either at one side of the margin (nine patients) or around the lesion (one patient); clustered spike sources were distant from the lesion in one patient. Among the patients with clustered spike sources near the lesion, further extensions (three patients) and distal scatters (three patients) were also observed. MEG spike sources were well lateralized and localized even in two patients without focal epileptiform discharges in the interictal scalp electroencephalography. Ten patients (77%) achieved Engel class I postsurgical seizure outcome. It is suggested that MEG is a safe and useful presurgical evaluation modality in pediatric patients with lesion localization-related epilepsy.

Signal characteristics of intraventricular electrodes recordings in human epilepsy: a case report

The case of a patient with intractable temporal lobe seizures and inadvertent unilateral intraventricular depth electrode placement is presented. The resting electroencephalograph (EEG) showed marked amplitude differences between the intraventricular electrode on the left and the parenchymal electrode on the right. All recorded seizures originated on the left side and in spite of its intraventricular location, frequency power spectra during the early ictal phase showed a marked increase in power for all frequency bands in the left depth electrode, exceeding that on the right. Analysis with Brain Electrical Source Analysis (BESA) software demonstrated marked ictal baseline shifts which were initially limited to the left side but changed to the right during clinical secondary generalization. In the immediate postictal state, all, except for infraslow, frequencies were markedly reduced in power. We conclude that intraventricular depth electrode contacts placed adjacent to the hippocampal structure can record interictal and ictal activity for all frequency bands, albeit at reduced amplitudes. Furthermore, infraslow activity can provide supplementary information about the epileptogenic zone.

Spatial MEG laterality maps for language: clinical applications in epilepsy

Functional imaging is increasingly being used to provide a noninvasive alternative to intracarotid sodium amobarbitol testing (i.e., the Wada test). Although magnetoencephalography (MEG) has shown significant potential in this regard, the resultant output is often reduced to a simplified estimate of laterality. Such estimates belie the richness of functional imaging data and consequently limit the potential value. We present a novel approach that utilizes MEG data to compute "complex laterality vectors" and consequently "laterality maps" for a given function. Language function was examined in healthy controls and in people with epilepsy. When compared with traditional laterality index (LI) approaches, the resultant maps provided critical information about the magnitude and spatial characteristics of lateralized function. Specifically, it was possible to more clearly define low LI scores resulting from strong bilateral activation, high LI scores resulting from weak unilateral activation, and most importantly, the spatial distribution of lateralized activation. We argue that the laterality concept is better presented with the inherent spatial sensitivity of activation maps, rather than being collapsed into a one-dimensional index.