Detection of recurrent activation patterns across focal seizures: Application to seizure onset zone identification
Introduction
The accurate identification of the epileptogenic zone in patients with medically refractory epilepsy is central to plan an efficient surgical intervention. Yet, after decades of surgical treatment experience, the outcome is not completely successful in a significant proportion of patients (Spencer and Huh, 2008) for several causes including complex epileptogenic networks, surgery technical limitations, among others. Nowadays invasive recording techniques such as stereoencephalography (SEEG) (Munari and Bancaud, 1985, Talairach et al., 1974, Guenot et al., 2002, Engel et al., 2005) provide a continuous monitoring tool for pre-surgical diagnosis that becomes specially effective in patients with challenging focal epilepsies. However, those patients exhibiting complex activation patterns across different seizures still represent a diagnostic challenge because they involve time-consuming EEG evaluations that may lead to inconclusive interpretations. Hence, the use of quantitative tools that assess the consistency of ictal patterns across recurrent seizures might contribute to objectively identify the epileptogenic zone, thus improving pre-surgical diagnosis (Engel et al., 2005) and reducing potential failures. In addition, quantitative measures might provide additional insights into the mechanisms of focal ictogenesis (Goodfellow et al., 2016) when correlated with available structural and functional information.
Over the last decade, several studies have proposed biomarkers to evaluate the epileptogenicity of recorded intracranial structures. A great number of them have analyzed spectral features of SEEG signals (Bartolomei et al., 2008, David et al., 2011, Gnatkovsky et al., 2011, Gnatkovsky et al., 2014, Andrzejak et al., 2015), which mimic the patterns that are visually identified as epileptogenic during diagnostic EEG inspection (Gnatkovsky et al., 2014, Lagarde et al., 2016). In these works, epileptogenic biomarkers are typically built around two variables, the activation of signal’s spectral properties (e.g., the relative amount of signal power in a frequency band (Bartolomei et al., 2008)) and the onset time of this activation, which measure respectively the amount of ictal activity and the degree of early participation of such structures during seizures. Epileptogenicity is then defined for each structure by combining both variables into an index that may be averaged across distinct seizures (Bartolomei et al., 2008, David et al., 2011, Andrzejak et al., 2015) to obtain a single biomarker per region. Critically, these mathematical transformations leading to a single value per region might prove inaccurate when averages are performed across seizures with heterogeneous activation patterns.
In the current study, we developed a flexible, robust and visual-friendly computer-aided method to assess the homogeneity of two ictal-driven channel spectral patterns (mean power change and power change onset time) across recurrent focal seizures. We evaluated the method in a group of seven epileptic patients with temporal lobe drug-resistant focal seizures. In all patients, we obtained seizure averages of both activation patterns for each recorded channel that were represented in two-dimensional plots for complementary clinical evaluation. As an example of application, we used each averaged pattern to characterize the seizure onset zone (SOZ) of seven patients with well identified seizure focus and tested the variability of the results obtained against the main method’s parameters. The proposed procedure may be integrated into existing epileptogenic indices by choosing the frequency band of interest and appropriately plugging in each averaged pattern, thus contributing to identify in a robust form central regions in the generation and spread of focal seizures.
Section snippets
Ethics statement
All diagnostic and surgical procedures were approved by The Clinical Ethical Committee of our Hospital.
Patients and recordings selection
A total number of 46 focal seizures from seven patients with pharmacoresistant epilepsy were analyzed. A summary of all patients’ characteristics is given in Table 1. Seizure onset and termination times of each seizure were independently marked by two epileptologists (RR and AP) using standard clinical assessment. For each seizure we analyzed SEEG recordings from the marked ictal epoch
Normalized seizure ensemble (NSE)
In all studied patients, we computed the channels’ nMA following the method described above. The homogeneity of normalized values across recurrent seizures was assessed in six patients (Patient 3 had only one seizure) prior to perform seizure averages. In these cases, we identified a NSE composed by a large proportion of focal seizures (, mean standard deviation, Table S1), which remained stable across a wide range of channel thresholding conditions (Fig. 3A). Furthermore, we could
Discussion
We proposed a robust methodology to assess recurrent ictal-driven patterns in intracranially recorded signals based on a temporal high-resolution spectral method (Hilbert transform Le Van Quyen et al., 2001, Oweis and Abdulhay, 2011). Our approach relies on finding an ensemble of focal seizures for every patient where these patterns can be compared across seizures and then, single biomarkers per channel can be meaningfully obtained to help define the most epileptogenic regions prior to surgery.
Conclusions
The present study proposes a general method to robustly quantify spectral activation patterns that are consistent across recurrent focal seizures. By exploiting the consistent channel activation profiles that patients exhibit across focal seizures, we propose a normalization procedure that allows to jointly analyze channel activation patterns from different seizures. The two-dimensional analysis of the time-average activation and the activation onset time during seizures provides a good
Disclosure
None of the authors has any conflict of interest to disclose. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.
Acknowledgments
The research was supported by the European Union's Horizon 2020 research and innovation program under Grant Agreement No. 720270. M.V. was supported by the Catalan Research project AL08814 - AGAUR - 2014SGR856. G.D. was supported by the European Research Council Advanced Grant DYSTRUCTURE (Grant 295129) and by the Spanish Research Project AEI/FEDER-PSI2016–75688-P. A.T.C. was supported by the European Research Council Advanced Grant DYSTRUCTURE (Grant 295129) and by the Spanish Research Project
References (24)
- et al.
How important is the seizure onset zone for seizure dynamics?
Seizure
(2015) - et al.
The identification of distinct high-frequency oscillations during spikes delineates the seizure onset zone better than high-frequency spectral power changes
Clin Neurophysiol
(2016) - et al.
Comparison of Hilbert transform and wavelet methods for the analysis of neuronal synchrony
J Neurosci Met
(2001) - et al.
Identification of seizure onset zone and preictal state based on characteristics of high frequency oscillations
Clin Neurophysiol
(2015) - et al.
Outcomes of epilepsy surgery in adults and children
Lancet Neurol
(2008) - et al.
Localization of epileptogenic zone on pre-surgical intracranial EEG recordings: toward a validation of quantitative signal analysis approaches
Brain Topogr
(2015) - et al.
Epileptogenicity of brain structures in human temporal lobe epilepsy: a quantified study from intracerebral EEG
Brain
(2008) - et al.
Stereo electroencephalography–guided radiofrequency thermocoagulation (SEEG-guided RF-TC) in drug-resistant focal epilepsy: results from a 10-year experience
Epilepsia
(2017) A power primer
Psychol Bull
(1992)- et al.
Imaging the seizure onset zone with stereo-electroencephalography
Brain
(2011)
Invasive recordings from the human brain: clinical insights and beyond
Nat Rev Neuro
Cited by (10)
Quantitative EEG biomarkers for epilepsy and their relation to chemical biomarkers
2021, Advances in Clinical ChemistryCitation Excerpt :This index ranges from 0 to 1 and indicates structures that are highly likely to generate seizures. Spectral activation patterns are higher for sensors in the seizure onset zone compared to sensors lying outside the seizure onset zone, with highest average effect size in alpha-beta bands (8–20 Hz) and best discrimination in the gamma band (20–70 Hz) [99]. Another study determined the ictally dominant frequency and found that the respective activity started on average prior to the onset of initial EEG signs determined by visual inspection, and that the frequency incremented during the seizure [100].
Epileptogenicity Mapping: A Quantitative Approach to Identify the Seizure Onset
2020, Neurosurgery Clinics of North AmericaCitation Excerpt :Numerous biomarkers were therefore proposed to characterize epileptogenicity based on selected spectral or time-frequency features. From the recent work of the group from Hospital del Mar in Barcelona, power spectral activation patterns in gamma band (20–70 Hz) were first studied, by which further enhanced data-driven method, based on the global spectral activation in different frequencies and the activation of entropy, was developed to find the temporal scale and frequency range of SOZ neural oscillations {VilaVidal:2019cr}.5 In the earlier work from the Milan group, a wider range of frequencies (1–250 Hz) was reported to achieve discrimination between SOZ and non-SOZ.6
Low entropy map of brain oscillatory activity identifies spatially localized events: A new method for automated epilepsy focus prediction
2020, NeuroImageCitation Excerpt :Based on previous studies (Bartolomei et al., 2008; David et al., 2011) we defined ictal-driven activity as an increase in signal power from pre-ictal to ictal epochs. In particular, we used the mean activation (MA) measure (Vila-Vidal et al., 2017), which quantifies the average spectral activation of each targeted brain structure for pre-defined frequency and time windows of interest (Fig. 1). However, while most studies constrained spectral activations to occur in preselected frequency bands, our method automatically infers the characteristic temporal scale and frequency range of locally enhanced oscillations in each case, thus maximizing the amount of information relevant for SOZ detection in a patient and seizure-specific context.
Classification of Epileptic Activity Through Temporal and Spatial Characterization of Intracranial Recordings
2020, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
- 1
These authors jointly supervised this work.