Epilepsy is one of the most common and debilitating neurological disorders in children (Fiest et al., 2017), with one-third of patients unresponsive to pharmacotherapy (Geerts et al., 2010; Berg et al., 2014). In focal epilepsy, seizures arise from one region of the brain, whose surgical resection provides effective treatment (Wiebe et al., 2001; Engel et al., 2012). Epilepsy surgery is now recognized as the first-line, standard of care in well-selected pediatric patients (Dwivedi et al., 2017) with a clear seizure onset zone (SOZ) as the site of the primary organization of ictal discharges (Mountz et al., 2017). Well-defined cases with clearly visible lesions on magnetic resonance imaging (MRI), such as tumors, cavernous malformations, or focal cortical dysplasia (FCD) type IIb, have an 85–95% chance of seizure freedom after surgery (Colombo et al., 2012; Englot et al., 2016). However, poorly-defined cases (PDCs), which can be MRI-negative or have ill-defined signal abnormalities on MRI, such as FCD type IIa remain challenging and present considerably less positive surgical outcome (Colombo et al., 2012; Wang et al., 2016). For such PDCs, multimodal neuroimaging and in some cases, an invasive intracranial recording is required to identify the origin of seizures (Mountz et al., 2017). Nevertheless, and despite advances in neuroimaging and electrophysiology methods epilepsy surgical outcomes have not improved substantially. More research is needed to build a deeper understanding of how focal epilepsy affects large scale neural dynamics and assist with the clinical decision making of delineating the SOZ.

Magnetoencephalography (MEG) is now recognized as a valuable asset in the armamentarium of epileptologists and researchers. It allows for time-resolved source imaging of interictal spike waveforms and slow/fast oscillatory components of brain activity (Mountz et al., 2017; Choi and Wang, 2020; Stefan and Rampp, 2020), with reduced dependence on head tissue characteristics with respect to electroencephalography (EEG; Baillet, 2017). While an increasing number of studies have investigated the value of ictal MEG (Alkawadri et al., 2018), routine clinical MEG has so far mainly been used in the context of interictal measurements (Stefan and Rampp, 2020). This is in part due to the practical implications imposed by the multi-hour recordings often required to capture ictal events. However, we have recently shown that ictal MEG studies, and particularly overnight ictal MEG for patients with nocturnal epilepsy, can be reliably conducted and have the potential to improve SOZ identification (Moreau et al., 2020).

There are currently no widely accepted guidelines for the analysis and interpretation of ictal MEG data. Ictal spikes could in principle be analyzed akin to interictal events, however, with limited ability to map the Spatio-temporal evolution of epileptic activity after seizure onset. Spectral measurements of ongoing MEG source brain activity can be resolved in time and do not depend on an expert marking of discrete events. They, therefore, have the potential to reveal early, anatomically specific changes of neural dynamics during seizure initiation. Power spectrum density (PSD) estimates of electrophysiological brain signals typically reveal narrowband peaks that correspond to oscillatory signal components, superimposed on a broadband aperiodic profile traditionally discarded as background or brain “noise” (Haller et al., 2018). Studies of narrowband neurophysiological signals and their amplitude changes have been extensive in epilepsy research (Jacobs et al., 2012; Schönherr et al., 2017; Sueri et al., 2018).

There has been renewed interest in the objective characterization of the aperiodic spectral component and its implications in possible biased interpretations of narrowband signal changes (He, 2014). The power of the background spectral component decreases with frequency, as typical of scale-free brain dynamics, following a power-law function of the form: Power ~ 1/f^β, with β a positive exponent (He, 2014). The aperiodic spectral component evolves with aging (Voytek et al., 2015) and neurodevelopment (He et al., 2019) and is affected in neurological disorders (Peterson et al., 2018) and by sedative drugs (Colombo et al., 2019). Yet, its functional significance concerning co-existing narrowband signals remains to be clarified. The 1/f^β component can readily be extracted from ongoing brain activity, without predefining frequency bands or specifying paroxysmal events (e.g., spikes). The profile of the aperiodic spectral component can be approximated from the parametric modeling of the PSD estimate derived from empirical signals. The two scalar parameters of the related parametric model are the spectral offset (y_offset) and the slope, represented by the exponent (β; Haller et al., 2018). The offset accounts for the overall magnitude of the aperiodic spectral component, and the exponent accounts for its curved shape. A broadband increase of higher frequency power and/or decrease of lower-frequency power leads to a decreased slope. New practical analytical tools are available to fit these model parameters from the signal PSD. The resulting model fit can thereafter be removed from the PSD to investigate the actual narrowband oscillatory activity, if remaining, without possible confounds from the broadband arrhythmic activity (Haller et al., 2018).

Computational modeling and empirical data indicate that changes in the aperiodic slope are related to the synchrony of cell firing in neuronal networks, with an increased slope associated with increased population synchrony (Freeman and Zhai, 2009; Voytek and Knight, 2015; Voytek et al., 2015). The offset parameter has also been related to the aggregate of the underlying local neuronal spiking activity (Manning et al., 2009; Miller et al., 2009, 2013; Voytek and Knight, 2015). It has also been suggested that changes in excitation and inhibition properties of cell assemblies can produce shifts in the PSD slope (Gao et al., 2017; Peterson et al., 2018). One possible mechanism of epilepsy is hypothesized to be related to periods of excessive discharges combined with synchronous excitation of neuronal populations (Scharfman, 2007). Historically, hypersynchrony and hyperexcitability have been related to an imbalance in regional excitation/inhibition (E/I) via increased excitatory neurotransmitter conductance and/or disruption of mechanisms inhibiting firing (Khazipov, 2016; Shao et al., 2019). However, this view has been challenged as too limited, with additional possible mechanistic contributors to epileptogenesis being attributed to E/I imbalance, such as with synchronization via inhibition (Shao et al., 2019).

Here, we report on time-evolving changes of aperiodic broadband brain activity and oscillatory components of the power spectrum in an ictal MEG recording of a patient with focal drug-resistant epilepsy.

An 8-year-old girl with focal drug-resistant epilepsy was admitted for presurgical evaluation. Based on prior routine EEG and imaging studies (MRI, PET, SPECT), seizures were suspected to localize to the right fronto-temporal cortex (Figures 1A–D). At the time of admission, the patient was having about 10 seizures per day, which were characterized by brief (up to ~10 s) periods of unresponsiveness and a blank expression. Early MRIs showed only subtle asymmetry in temporal cortex sulci. Routine EEG showed frequent seizures over right fronto-temporal electrodes, mostly during sleep. Given the frequent nighttime seizures, we brought in the patient for an overnight MEG recording (~4 h of MEG with simultaneous EEG). During this recording, one ictal event was recorded (Figure 1E). Electrographic changes on MEG and EEG preceded clinical onset by about 6 s. Clinically, the seizure was characterized by arousal followed by a forced head version to the left. Ictal MEG source imaging localized to the right anterior temporal cortex and was concordant with source localization of interictal MEG spikes as well as subtle blurring at the gray-white matter junction on a 3T T1/FLAIR MRI. Interictal FDG-PET and SPECT were also concordant, localizing to the right temporal cortex. The patient underwent a tailored right temporal resection from the anterior portion of the temporal neocortex and amygdala to the mid-posterior temporal neocortex but sparing the hippocampus proper. She continues to be seizure-free at 22 months follow-up. Histopathological assessment of dedicated surgical sub-specimens confirmed frank FCD type IIa in the MEG-defined SOZ. Further clinical details and surgical pathology results are presented in Moreau et al. (2020).

Preprocessing of the data and removal of bad segments resulted in ~9 min of data before seizure onset, one 5-s epoch moments before seizure onset, one 5-s ictal epoch, and a post-ictal epoch shortly after seizure resolution. The interictal recording contained one interictal spike and the patient was in sleep stage N3 for most of the recording and sleep stage N2 from ~1 min before seizure onset. The data immediately following the ictal epoch was not analyzable due to movement artifact.

This case report presents an exploratory analysis of unique overnight ictal MEG recordings in a child with drug-resistant epilepsy. The ictal MEG SOZ was delineated and concordant with the surgical resection and confirmed FCD type IIa pathology. The patient remains seizure-free 22 months post-surgery. We investigated the dynamic changes in both the aperiodic broadband signals and narrow-band peaks of the power spectrum.

To our knowledge, no prior studies have examined temporal changes and spatial differences in the aperiodic power spectrum components of ictal MEG data. We found that both the offset and exponent parameters of the aperiodic fit increased shortly before the seizure and reached their highest values mid-seizure. This increase was observed in both the SOZ and control ROIs. However, values were relatively higher in the SOZ, especially as compared to the contralateral ROI (nSOZ₃). We interpret these observations as possibly being related to rapid ictal spread to contiguous regions surrounding the SOZ (nSOZ₁ and nSOZ₂), which would not have impacted the contralateral hemisphere ROI (nSOZ₃). However, it must be kept in mind that the influence of lower frequency dynamics, which is represented in the aperiodic component is more widespread and can therefore influence contiguous areas.

These results are compatible with the theory suggesting that increased aperiodic slope is related to hypersynchrony of oscillatory coupling, resulting in organized spiking (Manning et al., 2009; Miller et al., 2009, 2013; Voytek and Knight, 2015). This could manifest via increased synchrony in the SOZ immediately before and during seizures. Increased aperiodic signal power (offset), may reflect increased population spiking activity (Manning et al., 2009; Miller et al., 2009, 2013; Voytek and Knight, 2015). Note that the offset and slope are expected to be correlated measures (Haller et al., 2018). We do not expect that these changes are related to changes in sleep stages, since a decrease in slope would be expected from a transition between deeper sleep (N3) to lighter sleep (N1, N2; Lendner et al., 2019; Miskovic et al., 2019).

Spectral analysis in canonical frequency bands of both the original and 1/f compensated PSD showed that changes in power over time of the original PSD were similar to the patterns observed in aperiodic component parameters. Indeed, delta and theta power values were higher in the SOZ compared to nSOZ₁ and this difference decreased from ~100 s before seizure onset. Furthermore, we did not observe clear differences between the SOZ and nSOZ₂ at any time point or for any frequency band. Lastly, differences between the SOZ and the contralateral ROI also increased from ~120 s before ictal onset, with a higher power in the SOZ expressed in all frequency bands while power values were higher in the delta, beta, and gamma bands less than 120 s before seizure onset in the contralateral ROI. These findings suggest that such differences might be due to broadband changes, and not solely to changes in narrow-band oscillatory components.

Indeed, removal of the 1/f components from the PSD confirmed this hypothesis: most spatial differences (i.e., differences between ROIs) and temporal changes that consisted of narrow-band components disappeared after 1/f compensation. The only difference that remained was between the SOZ and the contralateral ROI in the delta and theta frequency bands. Oscillatory signal power was higher in the SOZ in the theta frequency band and vice versa for the delta band, although less pronounced.

These results show promise for SOZ delineation without the need for the definition of discrete events related to seizure onset and frequency bands of interest by an expert.

Some limitations include the limited amount of peri-ictal recording that could be analyzed due to extensive head movement artifacts that occurred shortly following electrographic seizure onset. Inter-ictal and pre-ictal data segments were substantially longer, providing more data samples and potentially more robust outcomes. This case was meant to illustrate novel applications of methods to more precisely characterize the spectral components of ictal MEG recordings. These results will need to be reproduced and further investigated in larger case series and with longer ictal recordings to better define the range of observable dynamic changes that can occur in the course of seizures.

In this study, we provide early evidence suggesting that the broadband aperiodic component of the PSD contains useful neurophysiological information rather than just “background brain noise” and could contribute to helping define the SOZ. Studies with longer ictal MEG recordings will be needed to confirm and expand on the present findings. Overall, our data suggest that this novel approach holds promise for aiding in localizing the SOZ in children with focal epilepsy.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Research Ethics Board of the Montreal University Health Centre. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin. Written informed consent was obtained from the minor(s)’ legal guardian/next of kin for the publication of any potentially identifiable images or data included in this article.

SH conducted the analysis and interpretation of the data, drafted and revised the manuscript for intellectual content. JM and RD contributed to the design and conceptualization of the study, a major role in the acquisition of data, analysis, interpretation of the data, and revised the manuscript for intellectual content. ES-T helped with marking the seizure and interpreting the EEG. SA analyzed the pathology specimens used for confirmation of the seizure onset zone. SB contributed to the design and conceptualization of the study, analysis, interpretation of the data and revised the manuscript for intellectual content. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer GP declared a shared affiliation with the authors to the handling editor at time of review.