A glossary of terms most commonly used by clinical electroencephalographers : G. E. Chatrian (Chairman), L. Bergamini, M. Dondey, D. W. Klass, M. Lennox-Buchthal and I. Petersén

Extraoperative electrical cortical stimulation (ECS) facilitates defining the seizure onset zone (SOZ) and eloquent cortex. The clinical relevance of stimulation-induced afterdischarges (ADs) is not well defined.

Fifty-five patients who underwent intracranial electroencephalogram evaluations with ECS were retrospectively identified. ADs were identified in these recordings and categorized by pattern, location, and association with stimulation-induced seizures.

ADs were generated in 1774/9285 (19%) trials. Rhythmic spikes and irregular ADs within the stimulated bipolar contact pair were predictive of location within the SOZ compared to non-epileptogenic/non-irritative cortex (rhythmic spikes OR 2.24, p = 0.0098; irregular OR 1.39; p = 0.013). ADs immediately preceding stimulated seizures occurred at lower stimulation intensity thresholds compared to other stimulations (mean 2.94 ± 0.28 mA vs. 4.16 ± 0.05 mA respectively; p = 0.0068).

Changes in AD properties can provide clinically relevant data in extraoperative stimulation mapping.

Although not exclusive to the SOZ, the generation of rhythmic spikes may suggest that a stimulation location is within the SOZ, while decreased stimulation intensity thresholds eliciting ADs may alert clinicians to a heightened probability of seizure generation with subsequent stimulation.

This study develops new machine learning architectures that are more adept at detecting interictal epileptiform discharges (IEDs) in scalp EEG. A comparison of results using the average precision (AP) metric is made with the proposed models on two datasets obtained from Baptist Hospital of Miami and Temple University Hospital.

Applying graph neural networks (GNNs) on functional connectivity (FC) maps of different frequency sub-bands to yield a novel architecture we call FC-GNN. Attention mechanism is applied on a complete graph to let the neural network select its important edges, hence bypassing the extraction of features, a model we refer to as CA-GNN.

On the Baptist Hospital dataset, the results were as follows: Vanilla Self-Attention $\to 0.9029\pm$0.0431, Hierarchical Attention $\to 0.8546\pm$0.0587, Vanilla Visual Geometry Group (VGG) $\to 0.92\pm$0.0618, Satelight $\to 0.9219\pm$0.046, FC-GNN $\to 0.9731\pm$0.0187, and CA-GNN $\to 0.9788\pm$0.0125. In the same order, the results on the Temple University Hospital dataset are 0.9692, 0.9113, 0.97, 0.9575, 0.963, and 0.9879.

Based on the good results they yield, GNNs prove to have a strong potential in detecting epileptogenic activity.

This study opens the door for the discovery of the powerful role played by GNNs in capturing IEDs, which is an essential step for identifying the epileptogenic networks of the affected brain and hence improving the prospects for more accurate 3D source localization.

To evaluate the diagnostic accuracy of electromagnetic source imaging (EMSI) in localizing spikes and predict surgical outcome in children with drug resistant epilepsy (DRE) due to focal cortical dysplasia (FCD).

We retrospectively analyzed magnetoencephalography (MEG) and high-density (HD-EEG) data from 23 children with FCD-associated DRE who underwent intracranial EEG and surgery. We localized spikes using equivalent current dipole (ECD) fitting, dipole clustering, and dynamical statistical parametric mapping (dSPM) on EMSI, electric source imaging (ESI), and magnetic source imaging (MSI). We calculated the distance from the seizure onset zone (D_SOZ) and resection (D_RES). We estimated receiver operating characteristic (ROC) curves with Youden’s index (J) to predict outcome.

EMSI presented shorter D_SOZ (15.18 ± 9.06 mm) and D_RES (8.56 ± 6.24 mm) compared to ESI (D_SOZ: 25.04 ± 16.20 mm, p < 0.009; D_RES: 18.88 ± 17.30 mm, p < 0.03) and MSI (D_SOZ: 23.37 ± 8.98 mm, p < 0.03; D_RES: 15.51 ± 10.11 mm, p < 0.02) for clustering in patients with good outcome. Clustering showed shorter D_SOZ and D_RES compared to ECD fitting and dSPM (p < 0.05). EMSI had higher performance as outcome predictor (J = 70.63%) compared to ESI (J = 41.27%) and MSI (J = 33.33%) for clustering.

EMSI provides superior localization and improved predictive performance than individual modalities.

EMSI can help the surgical planning and facilitate the localization of epileptogenic foci.

We review here the neuronal mechanisms that cause seizures in focal epileptic disorders and, specifically, those involving limbic structures that are known to be implicated in human mesial temporal lobe epilepsy. In both epileptic patients and animal models, the initiation of focal seizures - which are most often characterized by a low-voltage fast onset EEG pattern - is presumably dependent on the synchronous firing of GABA-releasing interneurons that, by activating post-synaptic GABA_A receptors, cause large increases in extracellular [K⁺] through the activation of the co-transporter KCC2. A similar mechanism may contribute to seizure maintenance; accordingly, inhibiting KCC2 activity transforms seizure activity into a continuous pattern of short-lasting epileptiform discharges. It has also been found that interactions between different areas of the limbic system modulate seizure occurrence by controlling extracellular [K⁺] homeostasis. In line with this view, low-frequency electrical or optogenetic activation of limbic networks restrain seizure generation, an effect that may also involve the activation of GABA_B receptors and activity-dependent changes in epileptiform synchronization. Overall, these findings highlight the paradoxical role of GABA_A signaling in both focal seizure generation and maintenance, emphasize the efficacy of low-frequency activation in abating seizures, and provide experimental evidence explaining the poor efficacy of antiepileptic drugs designed to augment GABAergic function in controlling seizures in focal epileptic disorders.

Interictal activity and seizures are the hallmarks of focal epileptic disorders (which include mesial temporal lobe epilepsy, MTLE) in humans and in animal models. Interictal activity, which is recorded with cortical and intracerebral EEG recordings, comprises spikes, sharp waves and high-frequency oscillations, and has been used in clinical practice to identify the epileptic zone. However, its relation with seizures remains debated. Moreover, it is unclear whether specific EEG changes in interictal activity occur during the time preceding the appearance of spontaneous seizures. This period, which is termed “latent”, has been studied in rodent models of MTLE in which spontaneous seizures start to occur following an initial insult (most often a status epilepticus induced by convulsive drugs such as kainic acid or pilocarpine) and may mirror epileptogenesis, i.e., the process leading the brain to develop an enduring predisposition to seizure generation. Here, we will address this topic by reviewing experimental studies performed in MTLE models. Specifically, we will review data highlighting the dynamic changes in interictal spiking activity and high-frequency oscillations occurring during the latent period, and how optogenetic stimulation of specific cell populations can modulate them in the pilocarpine model. These findings indicate that interictal activity: (i) is heterogeneous in its EEG patterns and thus, presumably, in its underlying neuronal mechanisms; and (ii) can pinpoint to the epileptogenic processes occurring in focal epileptic disorders in animal models and, perhaps, in epileptic patients.