Interictal-Ictal Transitions
3.1) The Spatio-Temporal Evolution of EEG Correlation Clusters
Christian Rummel, Gerold Baier, Markus Müller
Facultad de Ciencias, Universidad Autónoma del Estado de Morelos, Avenida Universidad 1001, 62209 Cuernavaca, Morelos, México
Epileptic seizures represent changes in the correlation structure of brain activity. For a deeper understanding of seizure dynamics it is important to gain information about how strong different parts of the brain co-act in the respective phases. This information could be obtained from algorithms that were able to reliably detect the time evolution of correlation clusters. We present a multivariate approach to the detection of clusters in complex spatio-temporal systems that is based on eigenvectors of the equal-time correlation matrix. Instead of the largest eigenvectors themselves suitable linear combinations are exploited that are obtained systematically by maximization of a distance measure. These Cluster Participation Vectors (CPV) eliminate the disturbing effect of inter-cluster correlations and their components allow for improved conclusions on the involvement of channels in clusters. In model data the algorithm is able to correctly identify up to four clusters even in the presence of strong inter-cluster correlations and noise contamination. We demonstrate the usefulness of a running window application of the concept of CPV to EEG recordings at the example of standard surface EEG as well as ECoG with inter-ictal and ictal activity.
3.2) Correlation Changes During Ictal Activity
Gerold Baier (1), Markus Müller (1), Christian Rummel (1), Kaspar Schindler (2), Klaus Lehnertz (3)
(1) Faculty of Sciences, Autonomous University of Morelos, 62209 Cuernavaca, Morelos, Mexico; (2) Clinic for Neurology, University of Bern, Bern, Switzerland; (3) Department of Epileptology, University of Bonn, D-53105 Bonn, Germany
Epileptic activity in the EEG is often characterized by specific correlation patterns in part or all of the electrodes. Using a recently introduced correlation matrix formalism and measures from random matrix theory we investigate the multivariate correlation patterns before, during, and after ictal activity. Using the spectrum of relative eigenvalues and the individually unfolded P(s) distribution [1,2] we describe the correlation changes in data from 8x8 cortical grids placed over a frontal lobe seizure onset zone. It is found that the correlation patterns are in qualitative agreement with previous studies obtained with all intracranially implanted electrodes [3]. There is a strong correlation increase that accompanies the termination of the ictal activity (c.f. [3]) and which is carried over into the postictal period, predominantly in low-frequency components. During ictal activity, there are typical alterations of correlation decrease and correlation increase. The results are compared with the eigenvalue patterns of matrices composed of mutual information coefficients. The mutual information spectrum of eigenvalues characterises the onset of the ictal activity as a decrease in the subdelta components. The termination of the seizure is best marked by a strong increase in mutual information of components faster than 1 Hz. We discuss the differences between the two multivariate measures which may be due to nonlinear correlations present in the data. [1] M. Müller, Y. López, C. Rummel, G. Baier, A. Galka, U. Stephani, H. Muhle, Localized Short Range Correlations in the Spectrum of the Equal Time Correlation Matrix. Phys. Rev. E 74, 041119 (2006). [2] G. Baier, M. Müller, U. Stephani, H. Muhle, Characterizing Correlation Changes of Complex Pattern Transitions: the Case of Epileptic Activity. Phys. Lett. A 363, 290 (2007). [3] K. Schindler, H. Leung, C.E. Elger, K. Lehnertz, Assessing Seizure Dynamics by Analysing the Correlation Structure of Multichannel intracranial EEG. Brain, 130, 65 (2007).
3.3) Activation of the Thalamus and Deactivation of the Caudate Nucleus and Cortex Precede Generalized Cortical Epileptiform Discharges
Friederike Moeller (1), Hartwig R. Siebner (2,3), Stephan Wolff (4), Hiltrud Muhle (1), Rainer Boor (5), Oliver Granert (2), Olav Jansen (4), Ulrich Stephani (1,5), Michael Siniatchkin (1)
(1) Department of Neuropediatrics, University Hospital Schleswig Holstein, Campus Kiel, Germany; (2) Department of Neurology, University Hospital Schleswig Holstein, Campus Kiel, Germany; (3) NeuroImageNord, Hamburg – Kiel -Lübeck, Germany; (4) Section for Neuroradiology, University Hospital Schleswig Holstein, Campus Kiel, Germany; (5) Northern German Epilepsy Centre for children and adolescents, Raisdorf, Germany
The pathophysiology of generalized epileptiform discharges (GED) is not completely understood. Thalamus, basal ganglia, and neocortex have been implicated in the generation of GED, yet the specific role of each structure remains to be clarified. In children with idiopathic generalised epilepsy (IGE), we performed combined EEG functional MRI (fMRI) to identify GED-related changes in blood oxygen level-dependent (BOLD) signal in the striato-thalamo-cortical network. In the six patients with a sufficient number of GEDs during fMRI, within-subject analysis demonstrated BOLD signal changes that preceded the GED. An increase in BOLD signal in the medial thalamus started 6 seconds before the onset of the GED. Decreases in cortical BOLD signal were mainly found in frontoparietal cortical areas and the precuneus and started 6 to 3 seconds before the GED. All patients showed a decrease in BOLD signal in the head of the caudate nucleus. The onset of deactivation was, however, quite variable. The temporospatial pattern of BOLD signal changes suggests that GED on the cortical surface is preceded by a sequence of neuronal events in the thalamo-cortical-striatal network. First there is an increase in thalamic activity that starts approx 6 seconds prior to the GED followed by a deactivation of the cortex and caudate nucleus. We propose that these early changes in BOLD signal reflect changes in neuronal activity that contribute to the generation of interictal GED and might play a crucial role in the transition from a normal to a generalized hypersynchronous pattern of neuronal activity.
3.4) Estimating the Average Amount of Common Information in Scalp EEG Recordings Towards Preictal State Discrimination
Georgia E. Polychronaki (1), Periklis Ktonas (3), Stylianos Gatzonis (2,3), Anna Siatouni (2,3), Hara Tsekou (2), Maria Stefanatou (3), Damianos Sakas (2,3), Konstantina Nikita (1)
(1) Biomedical Simulations and Imaging Laboratory, School of Electrical and Computer Engineering, National Technical University of Athens; (2) Epilepsy Surgical Treatment Unit, Department of Neurosurgery, University of Athens, "Evangelismos" Hospital; (3) Greek Centre for Neurosurgical Research "Prof. Petros Kokkalis"
There is an interest in EEG synchronization studies, towards preictal vs interictal state discrimination and the ultimate goal of seizure prediction (Mormann et al., Clinical Neurophysiology 2005; 116:569-587). Mainly intracranial EEG recordings have been used for this purpose, although the use of surface EEG would be more realistic for therapeutic intervention in the future. In this work we propose a technique for estimating the common information between adjacent channels of scalp EEG recordings, in order to pursue preictal vs interictal state discrimination. We use Moddemeijer’s estimator (Moddemeijer, Internal report paper no. 1588, Twente University, Department of Electrical Engineering, The Netherlands, 1986) of the Average Amount of Mutual Information (AAMI) statistic, which quantifies the amount of common information between two random variables (i.e., two EEG channels). We analyse data from 3 patients undergoing presurgical evaluation for epilepsy surgery, all suffering from temporal lobe epilepsy. The analysis is performed on a desktop computer, using a time-windowing method in different EEG frequency bands, for at least 3 hours before and 1 hour after each of the 12 seizures analysed, wherever available. Results indicate that during the preictal period (first 20-min), in temporal EEG channels where the seizure is first observed, a statistically significant decrease or increase in AAMI values exists, compared to AAMI values during the interictal period (interval, with minimum duration of 30-min, taken from the EEG record at least 3 hours before and 1 hour after any seizure), in 11/12 (0.5 – 30 Hz), 10/12 (0.5 – 8 Hz), 11/12 (8 – 14 Hz), 11/12 (14 – 30 Hz) seizures. The observed decreases in AAMI are predominant, possibly reflecting specific electrode position relative to the epileptogenic area.
3.5) Long-Term Evaluation of Preictal State Identification by ECG Changes in Partial Epilepsy
Mario Valderrama (1), Vincent Navarro (1,2), Benoit Crepon (1,2), Michel Le Van Quyen (1)
(1) Cognitive Neurosciences and Brain Imaging, LENA, CNRS UPR 640, Hôpital de la Pitié-Salpêtrière, Paris, France; (2) Epileptology Unit, Hôpital de la Pitié-Salpêtrière, Paris, France
Objectives: Long-term interactions between epileptic discharges and the neuroautonomic regulation of the heart are studied. Depending on their localisation, epileptic focus may be highly related to autonomic control centers in the brain so an epileptic activity could influence the autonomic system behaviour. We focus our interest on specific, autonomic changes which could be product of an epileptic activity during inter-ictal, pre-ictal and ictal states. Additionally, we determine the existence of particular signatures of impending seizures presented in the autonomic regulation the heart, useful for anticipation purposes. Materials: Analyses are performed on long-term intracranial-EEG/ECG recordings of 5 patients with refractory partial epilepsy and evaluated for up to 7-14 days. Methods: Autonomic activity is assessed through oscillations from R-R intervals of ECG using time-frequency analysis. Discrete Windowed Fourier Transform (DWFT) and Wavelet Transform (WT) are used to determine the influence of epileptic discharges on the autonomic activity. Statistical analysis is preformed in order to evaluate the existence of relevant information relating specific epileptic and autonomic activities. Results: Significant changes in autonomic activity can be detected during recurrent, specific-located epileptic discharges. Furthermore, patient-specific recurrent patterns preceding seizures are observed, indicating a possible pre-ictal signature of the autonomic system activity. Our results suggest that an on-line version of the analyses, trained on each patient's peri-ictal ECG, could serve as a basis for a seizure alarm system.
3.6) Long-Range Dependence of Epileptic Seizures
Bela Weiss (1,2), Zsuzsanna Vago (1), Ronald Tetzlaff (2), Tamas Roska (1,3)
(1) Faculty of Information Technology, Peter Pazmany Catholic University, Budapest, Hungary; (2) Institute of Applied Physics, Johann Wolfgang Goethe University, Frankfurt/Main, Germany; (3) Computer and Automation Research Institute, Hungarian Academy of Sciences, Budapest, Hungary
Considering spike trains it was found that there are long term correlations among interspike intervals. A fractal spike train process is statistically self-similar, which means that fluctuations and other properties over brief times are proportional to those measured over a longer period. EEG signals arise from cellular activities and can be considered as top of the hierarchy above field potentials. Due to this, we assumed that signals recorded by subdural and intracerebral macro electrodes keep the aforementioned properties at least for short intervals. These can be considered stationary. Fitting of self-similar process to EEG time series can provide a very concise description of the system. The self-similarity parameter - the Hurst exponent (HE) - is a hidden parameter. It describes the long-range dependence of the process. We implemented a method based on the rescaled adjusted range or R/S statistic for estimation of the HE. Preliminary analysis on patients suffering from temporal lobe epilepsy (21-patient database from Epilepsy Center Freiburg) showed that HE produces distinct changes during seizures. In this recent work we assessed different types of seizures from the same database and from National Institute of Neurosurgery, Budapest, Hungary. Other types of seizures can be also detected as well as temporal lobe seizures. In the preictal interval a gradual increase, in the postictal state a decrease of the HE can be observed for several seizures. This can provide more accurate modeling of EEG signals and the dynamics of epileptic seizures using stochastic processes.
3.7) Increasing Trends of Phase Synchrony and Correlation of Microwire Channels before Seizure Onset
Sanqing Hu (1), Matt Stead (1), Brian Litt (2), Rick Marsh (1), Greg Worrell (1)
(1) Mayo Clinic, Department of Neurology, Rochester, MN 55905; (2) University of Pennsylvania, Department of Neurology, Philadelphia, PA, 19104
RATIONALE: A hypothesis for focal seizure generation involves the progressive coalescence of microdomain islands of seizure activity. Study of quantitative metrics such as phase synchrony and correlation between microwire channels before macroscale seizure onset is therefore of fundamental interest. METHODS: We studied 10 patients with custom hybrid depth and subdural electrodes containing arrays of microwires and clinical macroelectrodes. The EEG was acquired using a broadband amplifier operating in parallel with the clinical EEG acquisition system. Phase synchrony and correlation between microwire electrodes was investigated. RESULTS: Broadband measures of phase synchrony and correlation between microwire channels show variable behavior prior to macroseizure onset. However, in some patients the phase synchrony and correlation between high-pass filtered (>70Hz) microwire data increases seconds before macroseizure onset. CONCLUSIONS: Increasing trends of phase synchrony and correlation between microwires before macroscale seizure onset supports a model of seizure generation involving the coalescence of microdomain islands. We speculate that the trend in phase synchronization and correlation between microdomains continues to some threshold value, beyond which macroscale seizure occurs. We demonstrate that neuronal oscillations and seizures occur over a wide range of spatial and temporal scales, and that microseizures are likely important in the generation of macroscale seizures.
3.8) Analysis of Activity Flows During Preictal and Ictal Periods
Anna Korzeniewska (1), Christophe C. Jouny (1), Rafal Kus (2), Nathan E. Crone (1), Gregory K. Bergey (1), Piotr J. Franaszczuk (1)
(1) Department of Neurology, Johns Hopkins University School of Medicine, 600 N. Wolfe St., Meyer 2-147, Baltimore, MD 21287, USA; (2) Department of Biomedical Physics, Institute of Experimental Physics, Warsaw University, ul. Hoza 69, 00-681 Warsaw, Poland
Better understanding and description of processes leading to onset and spread of epileptic seizure may help in prevention or early termination of seizure. Multichannel methods provide tools for analysis of interactions between different regions of brain but they often are not well suited for nonstationary epileptiform signals. In many epileptic patients, seizures originating from the same focus often produce very similar EEG signals, particularly recorded intracranially. Using information from multiple seizures improves the statistical properties of estimators, allowing for analysis of shorter stationary window. The short-time direct directed transfer function (SdDTF) method, recently developed to investigate the directions and intensities of activity flow between cortical regions during cognitive tasks with multiple trials, was applied to intracranial recordings of several patients with multiple seizures (5-80). A multichannel autoregressive model (MVAR) was fitted simultaneously to all recorded seizures. The seizures were aligned according to their ictal onsets, judged by visual inspection. Preictal intervals of 60 sec, as well as ictal segments of 60 sec were analyzed. The SdDTF method showed flow of activity from the ictal onset zone. During the preictal period significant flows of low frequency activity could be observed in the vicinity of the focus. These flows may reflect underlying processes of synchronization leading to seizure onset. Supported by: NINDS R01 NS40596 and NS48222
3.9) Spatial and Temporal Identification of Seizure Precursor Dynamics using a Phase Modeling Approach
Tobias Wagner (1,2), Hannes Osterhage (1,2), Christian E. Elger (1), Klaus Lehnertz (1,2,3)
(1) Department of Epileptology, University of Bonn; (2) Helmholtz-Institute for Radiation and Nuclear Physics, University of Bonn; (3) Interdisciplinary Center for Complex Systems, University of Bonn
Recent studies have indicated a significant predictive performance for bivariate EEG analysis techniques, allowing one to discriminate the pre-ictal from the inter-ictal period above chance level. In addition, some studies reported that the site selected as best for prediction was not in close vicinity to the epileptic focus but could be located in remote or even contralateral brain structures. This seemingly counterintuitive finding may indicate the importance of brain outside of the ictal-onset zone but within the “epileptic network” in generating clinical seizures. Addressing this issue we studied directional relationships – in the driver-responder sense – in multi-day, multi-channel invasive EEG recordings using a phase modeling approach (Rosenblum & Pikovsky, Phys Rev E 64, 045202, 2001). Inter-ictally, we observed two distinct regions that exhibit a pronounced dynamical dependence on the surrounding brain regions both on the side of the epileptic focus and in the opposite hemisphere. Using a priori knowledge as to the localization of the epileptic focus we were able to assign the focal region to one of these structures, which, however, is driven by surrounding brain regions. Interestingly, the other structure, which is a driving structure and is located in almost homologous contralateral brain regions, exhibited dynamical aspects that allowed us to discriminate the pre-ictal from the inter-ictal period at a high performance. If proven significantly, our findings indicate the high relevance of brain structures outside the epileptic focus in ictogenesis. Measuring directionality in multi-channel EEG recordings may help to identify target brain structures with potential precursor dynamics, particularly in prospective seizure prediction studies.
3.10) Statistical Evaluation of Measures of Scalar Time Series in Discriminating Preictal EEG States
Dimitris Kugiumztis (1), Angeliki Papana (1), Ioannis Vlachos (1), Paal G. Larsson (2)
(1) Department of Mathematical, Phsyical and Computational Sciences, Aristotle University of Thessaloniki, Thessaloniki 54124, Greece; (2) Department of Neurodiagnostics (SSE), Neuroclinic, Rikshospitalet, Baerum N-1306, Norway
We consider a large number of measures used in the statistical and nonlinear analysis of univariate time series and EEG in particular. We include measures of correlation (autocorrelation, bicorrelation, mutual information), entropy, spectral energy, complexity (largest Lyapunov exponent, algorithmic complexity, Hurst exponents), dimension (point density, false nearest neighbors), and goodness of fit from linear and nonlinear models. In addition, the following feature time series are extracted from the oscillations of each EEG channel: local minima and maxima and the time between them, minimum to maximum magnitude difference and interspike intervals. The measures are applied to the original and the feature time series for selected values of the method-specific parameters, giving a total of 284 measures. The measures are estimated on subsequent segments of 30sec of multichannel EEG at preictal stage of several hours as well as interictal stage. The objective of this study is two-fold: a) to assess whether some of the measures can discriminate between early and late preictal stages, and b) to find the measures with the best discriminating power. For this, statistical analysis of the measure values grouped in different preictal stages has been applied using receiver operating characteristic (ROC) curves and statistical testing. The results on 10 epileptic EEG records show that simple measures have the same, and at cases better, power in discriminating preictal stages than other more sophisticated measures.
3.11) Microanatomy of Epileptiform Activity in Human Multielectrode Recordings
C. A. Schevon (1), S. K. Ng (1), J. Cappell (2), R. R. Goodman (3), G. McKhann Jr (3), A. Waziri (3), A. Branner (4), A. A. Sosunov (3), F. Gilliam (1), C. E. Schroeder (5), R. G. Emerson (1,2)
(1) Columbia University, Dept of Neurology; (2) Columbia University, Dept of Pediatrics; (3) Columbia University, Dept of Neurological Surgery; (4) Cyberkinetics Neurotechnology Systems; (5) Columbia University, Dept of Psychiatry
Uncovering the role of microcircuits in seizure origin and development requires fine-scale observation of neuronal activity in the ictal region. We report the use of a dense 2D microelectrode array (MEA) to provide new details of interictal and ictal electrophysiological disturbances in human epileptogenic cortex in vivo. A 4mm square MEA (96 microelectrodes in a 10 x 10 grid, 400 micron spacing, 1 mm long, 3-5 micron diameter recording tips) (NeuroPort™ , Cyberkinetics Neurotechnology Systems, Foxboro, MA) was implanted in six patients with medically intractable focal epilepsy undergoing intracranial EEG (iEEG) monitoring at the Columbia University Medical Center. Chronic recordings of 2 – 14 days were obtained. In five of the six patients, the interictal µEEG revealed highly focal waveforms that resembled conventional epileptiform discharges (“focal µEDs”) or electrographic seizures (“micro-ictal appearing discharges”, or µIADs). These features were topographically restricted to areas spanning 0.2 to 4 mm2 and were not evident in the iEEG. Additionally, the presence of focal µEDs and µIADs correlated with the location of the MEA with respect to the epileptogenic zone. These early findings suggest that µEDs and µIADs are specific markers of epileptogenic cortex, suggesting a structure of sparsely distributed tiny epileptogenic foci, and that they may serve as precursors to seizures. The interictal-ictal transitions seen in two of the patients suggest that µIADs play a prominent role in the development and propagation of seizures.