The “Connectivity Epileptogenicity Index ” (cEI), a method for mapping the different seizure onset patterns in StereoElectroEncephalography recorded seizures
Introduction
Epilepsy surgery can significantly improve the quality of life of patients with drug-resistant epilepsy. It consists in the removal of the brain regions generating seizures. These regions, (called the “Epileptogenic zone ” (Bancaud, 1980)) must be thus precisely defined from anatomo-electro-clinical observations obtained during pre-surgical evaluation. In a large number of cases, this evaluation involves invasive EEG recordings, in particular stereo-electroencephalography (SEEG) (Bancaud and Talairach, 1965, Cardinale et al., 2016, Isnard et al., 2018).
Defining the epileptogenic zone from SEEG can be challenging (Bartolomei et al., 2017, Bartolomei et al., 2018). The epileptogenic zone is increasingly recognized as a network of hyperexcitable connected regions generating seizures and secondary leading to ictal spreading in propagation networks (Bartolomei et al., 2017). The electrical onset of seizures may appear quite complex in its aspect, both in term of involved frequencies and in term of spatial extension. In 75% of cases, the seizure onset patterns involve high frequencies, most often in the high beta or gamma band (Perucca et al., 2014, Lagarde et al., 2019).
In this context, methods for quantifying the epileptogenic zone have been developed over the past ten years in order to complement the SEEG interpretation (for review see (Andrzejak et al., 2015, Bartolomei et al., 2017)). These are based on a spectral analysis of the SEEG signal obtained in each region, using energy in the high frequency (beta-gamma bands), potentially combined with other measures (low frequency deactivation, preictal activity). The first method that was developed, and the one with the largest patient population published until now, is the Epileptogenicity Index (EI) (Bartolomei et al., 2008). EI combines the ratio of fast frequencies (beta/gamma) relative to slower frequencies and the time of involvement of each region. This method inspired a number of studies aimed at identifying the epileptogenic zone from SEEG recordings (David et al., 2011, Gnatkovsky et al., 2011, Gnatkovsky et al., 2014, Grinenko et al., 2018). However, as most of these methods are based on the detection/mapping of high frequencies, they are inefficient to detect slower patterns of onset that account for 20–30% of commonly observed SEEG patterns (Singh et al., 2015, Lagarde et al., 2019). This is a clear a limitation since seizures with slow onset have worse prognosis for surgery outcome than seizures with fast activity at onset (Singh et al., 2015, Lagarde et al., 2019). It has been suggested that seizures with slower patterns are more distributed and that these slow patterns are related to etiology - in particular type 1 focal cortical dysplasia (Lagarde et al., 2019). It should be noted that the EI cannot be modified to measure low-frequency onsets because these serve as the denominator of the energy ratio from which it is calculated.
In this context, other methods based on functional connectivity have been proposed to study focal seizures onset from intracranial recordings (Brazier, 1969, Lopes da Silva et al., 1989, Gotman and Levtova, 1996, Bartolomei et al., 2004, van Mierlo et al., 2013, Courtens et al., 2016). These approaches measure linear or nonlinear properties of the relationships between SEEG signals, during the seizure onset period or the transition from interictal to ictal periods. Thus, the functional connectivity between SEEG signals tends to be maximal before the emergence of the fast discharges and to decrease just after, followed by later increase during the seizure course (Courtens et al., 2016). Some studies have used measure of causality (or directed or effective connectivity) and have shown that the definition of ”leader/driver” regions may add important clues into the definition of the network generating seizures (Bartolomei et al., 2004, Varotto et al., 2012, van Mierlo et al., 2013, Courtens et al., 2016).
The epileptogenicity of a brain network (i.e. its ability to generate seizures) is related to the level of excitability and the level of connectivity (Proix et al., 2014, Hebbink et al., 2017). In the present study we therefore propose an approach which combines two types of measures in the same quantity. A measure of connectivity and a measure of the capacity of regions to generate seizures with a short delay (EI). The objective is to improve sensitivity while better accounting for the diversity of seizure onset patterns, including slower ones. The connectivity measure was based on the nonlinear correlation coefficient (h2) and on graph theory measures (out-degrees) (Courtens et al., 2016). We focused our study on a cohort of patients in whom the seizure onset patterns (SOP) have been previously defined (Lagarde et al., 2016), relatively homogeneous in terms of etiology (malformation of cortical development) and included a proportion of slow onset patterns (20%).
Section snippets
Patient selection and data collection
We retrospectively analyzed data from 51 patients with pharmacoresistant focal epilepsy, who underwent SEEG in the context of their presurgical evaluation in the Epilepsy Unit of the Clinical Neurophysiology Department in Timone Hospital, Marseille, France from 2000 to 2016. The 51 patients are extracted from a cohort of 53 patients with a malformation of cortical development (focal cortical dysplasia type 1 or 2 (FDC1 or 2), neurodevelopmental tumour, NDT), already studied in terms of seizure
Results
Data from 51 patients (27 female) were analyzed. Since the objective was not to measure intra-patient reproducibility but to compare two measures estimating SOZ in relation to visual analysis, we chose a single representative seizure per patient. We took the first recorded seizures with well-defined seizure onset pattern category. Table 1 summarizes clinical data. The mean age at epilepsy onset was 7.3 years (range 0–28); the mean age at SEEG acquisition was 23.5 years (range 2.75–56) and the
Discussion
In this study, we propose a new marker of the epileptogenic zone (cEI, Connectivity Epileptogenicity Index) based on the sum of a graph measure and a local estimate of the high frequency discharge. The cEI combines a graph connectivity measure (“out-degree”) allowing the evaluation of the leading nodes in the network (Courtens et al., 2016), and the epileptogenicity index (Bartolomei et al., 2008) acting as a quantified estimate of the rapid discharge at seizure onset. The main objective was to
Funding
Alexandra Balatskaya‘s research work was funded by an EAN fellowship 2019.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
References (30)
- et al.
Interpretation of SEEG recordings
Neurophysiol Clin
(2018) - et al.
Pre-ictal synchronicity in limbic networks of mesial temporal lobe epilepsy
Epilepsy Res
(2004) - et al.
AnyWave: a cross-platform and modular software for visualizing and processing electrophysiological signals
J Neurosci Methods
(2015) - et al.
Amygdala-hippocampus relationships in temporal lobe seizures: a phase coherence study
Epilepsy Res
(1996) - et al.
French guidelines on stereoelectroencephalography (SEEG)
Neurophysiol Clin
(2018) - et al.
EpiTools, A software suite for presurgical brain mapping in epilepsy: Intracerebral EEG
J Neurosci Methods
(2018) - et al.
Epileptogenic networks of type II focal cortical dysplasia: a stereo-EEG study
Neuroimage
(2012) - et al.
Interpretation of interdependencies in epileptic signals using a macroscopic physiological model of EEG
Clin Neurophysiol
(2001) - et al.
Localization of epileptogenic zone on pre-surgical intracranial EEG recordings: toward a validation of quantitative signal analysis approaches
Brain Topogr
(2015) Surgery of epilepsy based on stereotactic investigations–the plan of the SEEG investigation
Acta Neurochir Suppl (Wien)
(1980)
La stéréo-électroencéphalographie dans l'épilepsie: informations neurophysiopathologiques apportées par l'investigation fonctionnelle stéreotaxique
Epileptogenicity of brain structures in human temporal lobe epilepsy: a quantified study from intracerebral EEG
Brain
Defining epileptogenic networks: Contribution of SEEG and signal analysis
Epilepsia
Analysis of EEG activity recorded from electrodes implanted in deep structures of the human brain
Electroencephalogr Clin Neurophysiol
Implantation of stereoelectroencephalography electrodes: a systematic review
J Clin Neurophysiol
Cited by (21)
Personalized whole brain modeling of status epilepticus
2023, Epilepsy and BehaviorPersonalised virtual brain models in epilepsy
2023, The Lancet NeurologyRecent developments in stereo electroencephalography monitoring for epilepsy surgery
2022, Epilepsy and BehaviorCitation Excerpt :Moreover, this connectivity pattern may remain stable over days, irrespective of the ASM dose and clinical seizure burden [61]. Balatskaya reported a method of connectivity EI (cEI) analysis, which combines directed connectivity measures (based on nonlinear correlation coefficient and graph theory measures) with EI [62]. The cEI provided a better estimation of slow seizure-onset patterns than EI.
Localization of seizure onset zone with epilepsy propagation networks based on graph convolutional network
2022, Biomedical Signal Processing and ControlEpileptogenic networks in drug-resistant epilepsy with amygdala enlargement: Assessment with stereo-EEG and 7 T MRI
2022, Clinical NeurophysiologyCitation Excerpt :The amygdala was one of the core epileptogenic structures within the distributed epileptogenic network that preferentially involved temporal (mesial, lateral, or both), insular, ventro-lateral prefrontal, ventro-mesial prefrontal regions and pulvinar. Based on the Connectivity Epileptogenicity index(Balatskaya et al., 2020), we could identify four distinct patterns of epileptogenic network organization, with the temporo-insular type being the most common, encountered in half of cases, followed by equally represented temporo-insular-prefrontal, prefrontal-temporal mesial, and mesial temporal types. It is likely that our selection of drug-resistant cases with SEEG induced a bias in the selection of more complex cases.
The accuracy of quantitative EEG biomarker algorithms depends upon seizure onset dynamics
2021, Epilepsy ResearchCitation Excerpt :Doing so required us to account for the different seizure onset patterns. Previous work has attempted to improve the performance of quantitative biomarkers by calculating indices based on different features of the signal or by combining a variety of different indices, each of which is derived from different features of the EEG signal (Balatskaya et al., 2020; Gnatkovsky et al., 2014). The general strategy is to apply tools that are best suited for each individual seizure.
- 1
Equal contribution.