Summary: Purpose: Photosensitive epilepsy (PSE) is the most common form of reflex epilepsy presenting with electroencephalography (EEG) paroxysms elicited by intermittent photic stimulation (IPS). To investigate whether the neuronal network undergoes dynamic changes before and during the transition to an EEG epileptic discharge, we estimated EEG connectivity patterns in photosensitive (PS) patients with idiopathic generalized epilepsy.Methods: EEG signals were evaluated under resting conditions and during 14 Hz IPS, a frequency that consistently induces photoparoxysmal responses (PPRs) in PS patients. Partial directed coherence (PDC), a linear measure of effective connectivity based on multivariate autoregressive models, was used in 10 PS patients and 10 controls. Anterior versus posterior (F3, F4, C3, C4, and P3, P4, O1, O2) and interhemispheric connectivity patterns (F4, C4, P4, O2, and F3, C3, P3, O1) were estimated with focus on beta and gamma band activity.Key Findings: PDC analysis revealed an enhanced connectivity pattern in terms of both the number and strength of outflow connections in the PS patient group. Under resting condition, the greater connectivity in the PS patients occurred in the beta band, whereas it mainly involved the gamma band during IPS (i.e., the frequencies ranging from 40–60 Hz that include the higher harmonics of the stimulus frequency). Both at rest and during IPS, the differences between the PS patients and controls were due primarily to clearly increased connectivity involving the anterior cortical regions.Significance: Our findings indicate that PS patients are characterised by abnormal EEG hyperconnectivity, primarily involving the anterior cortical regions under resting conditions and during IPS. This suggests that, even if the occipital cortical regions are the recipient zone of the stimulus and probably hyperexcitable, the anterior cortical areas are prominently involved in generating the hypersynchronization underlying the spike‐and wave discharges elicited by IPS.

In the context of focal drug-resistant epilepsies, the surgical resection of the epileptogenic zone (EZ), the cortical region responsible for the onset, early seizures organization and propagation, may be the only therapeutic option for reducing or suppressing seizures. The rather high rate of failure in epilepsy surgery of extra-temporal epilepsies highlights that the precise identification of the EZ, mandatory objective to achieve seizure freedom, is still an unsolved problem that requires more sophisticated methods of investigation. Despite the wide range of non-invasive investigations, intracranial stereo-EEG (SEEG) recordings still represent, in many patients, the gold standard for the EZ identification. In this contest, the EZ localization is still based on visual analysis of SEEG, inevitably affected by the drawback of subjectivity and strongly time-consuming. Over the last years, considerable efforts have been made to develop advanced signal analysis techniques able to improve the identification of the EZ. Particular attention has been paid to those methods aimed at quantifying and characterising the interactions and causal relationships between neuronal populations, since is nowadays well assumed that epileptic phenomena are associated with abnormal changes in brain synchronisation mechanisms, and initial evidence has shown the suitability of this approach for the EZ localisation. The aim of this review is to provide an overview of the different EEG signal processing methods applied to study connectivity between distinct brain cortical regions, namely in focal epilepsies. In addition, with the aim of localizing the EZ, the approach based on graph theory will be described, since the study of the topological properties of the networks has strongly improved the study of brain connectivity mechanisms.

Although several brain regions show significant specialization, higher functions such as cross-modal information integration, abstract reasoning and conscious awareness are viewed as emerging from interactions across distributed functional networks. Analytical approaches capable of capturing the properties of such networks can therefore enhance our ability to make inferences from functional MRI, electroencephalography and magnetoencephalography data. Graph theory is a branch of mathematics that focuses on the formal modelling of networks and offers a wide range of theoretical tools to quantify specific features of network architecture (topology) that can provide information complementing the anatomical localization of areas responding to given stimuli or tasks (topography). Explicit modelling of the architecture of axonal connections and interactions among areas can furthermore reveal peculiar topological properties that are conserved across diverse biological networks, and highly sensitive to disease states. The field is evolving rapidly, partly fuelled by computational developments that enable the study of connectivity at fine anatomical detail and the simultaneous interactions among multiple regions. Recent publications in this area have shown that graph-based modelling can enhance our ability to draw causal inferences from functional MRI experiments, and support the early detection of disconnection and the modelling of pathology spread in neurodegenerative disease, particularly Alzheimer's disease. Furthermore, neurophysiological studies have shown that network topology has a profound link to epileptogenesis and that connectivity indices derived from graph models aid in modelling the onset and spread of seizures. Graph-based analyses may therefore significantly help understand the bases of a range of neurological conditions. This review is designed to provide an overview of graph-based analyses of brain connectivity and their relevance to disease aimed principally at general neuroscientists and clinicians.

Aim:In neuroscience research, data are quite often characterized by an imbalanced distribution between the majority and minority classes, an issue that can limit or even worsen the prediction performance of machine learning methods. Different resampling procedures have been developed to face this problem and a lot of work has been done in comparing their effectiveness in different scenarios. Notably, the robustness of such techniques has been tested among a wide variety of different datasets, without considering the performance of each specific dataset. In this study, we compare the performances of different resampling procedures for the imbalanced domain in stereo-electroencephalography (SEEG) recordings of the patients with focal epilepsies who underwent surgery.Methods:We considered data obtained by network analysis of interictal SEEG recorded from 10 patients with drug-resistant focal epilepsies, for a supervised classification problem aimed at distinguishing between the epileptogenic and non-epileptogenic brain regions in interictal conditions. We investigated the effectiveness of five oversampling and five undersampling procedures, using 10 different machine learning classifiers. Moreover, six specific ensemble methods for the imbalanced domain were also tested. To compare the performances, Area under the ROC curve (AUC), F-measure, Geometric Mean, and Balanced Accuracy were considered.Results:Both the resampling procedures showed improved performances with respect to the original dataset. The oversampling procedure was found to be more sensitive to the type of classification method employed, with Adaptive Synthetic Sampling (ADASYN) exhibiting the best performances. All the undersampling approaches were more robust than the oversampling among the different classifiers, with Random Undersampling (RUS) exhibiting the best performance despite being the simplest and most basic classification method.Conclusions:The application of machine learning techniques that take into consideration the balance of features by resampling is beneficial and leads to more accurate localization of the epileptogenic zone from interictal periods. In addition, our results highlight the importance of the type of classification method that must be used together with the resampling to maximize the benefit to the outcome.