Using bivariate signal analysis to characterize the epileptic focus: the benefit of surrogates.

The disease epilepsy is related to hypersynchronous activity of networks of neurons. While acute epileptic seizures are the most extreme manifestation of this hypersynchronous activity, an elevated level of interdependence of neuronal dynamics is thought to persist also during the seizure-free interval. In multichannel recordings from brain areas involved in the epileptic process, this interdependence can be reflected in an increased linear cross correlation but also in signal properties of higher order. Bivariate time series analysis comprises a variety of approaches, each with different degrees of sensitivity and specificity for interdependencies reflected in lower- or higher-order properties of pairs of simultaneously recorded signals. Here we investigate which approach is best suited to detect putatively elevated interdependence levels in signals recorded from brain areas involved in the epileptic process. For this purpose, we use the linear cross correlation that is sensitive to lower-order signatures of interdependence, a nonlinear interdependence measure that integrates both lower- and higher-order properties, and a surrogate-corrected nonlinear interdependence measure that aims to specifically characterize higher-order properties. We analyze intracranial electroencephalographic recordings of the seizure-free interval from 29 patients with an epileptic focus located in the medial temporal lobe. Our results show that all three approaches detect higher levels of interdependence for signals recorded from the brain hemisphere containing the epileptic focus as compared to signals recorded from the opposite hemisphere. For the linear cross correlation, however, these differences are not significant. For the nonlinear interdependence measure, results are significant but only of moderate accuracy with regard to the discriminative power for the focal and nonfocal hemispheres. The highest significance and accuracy is obtained for the surrogate-corrected nonlinear interdependence measure.

Internetwork and intranetwork communications during bursting dynamics: applications to seizure prediction.

We use a simple dynamical model of two interacting networks of integrate-and-fire neurons to explain a seemingly paradoxical result observed in epileptic patients indicating that the level of phase synchrony declines below normal levels during the state preceding seizures (preictal state). We model the transition from the seizure free interval (interictal state) to the seizure (ictal state) as a slow increase in the mean depolarization of neurons in a network corresponding to the epileptic focus. We show that the transition from the interictal to preictal and then to the ictal state may be divided into separate dynamical regimes: the formation of slow oscillatory activity due to resonance between the two interacting networks observed during the interictal period, structureless activity during the preictal period when the two networks have different properties, and bursting dynamics driven by the network corresponding to the epileptic focus. Based on this result, we hypothesize that the beginning of the preictal period marks the beginning of the transition of the epileptic network from normal activity toward seizing.

Dissecting out conscious and unconscious memory (sub)processes within the human medial temporal lobe.

The human medial temporal lobe (MTL) system mediates memories that can be consciously recollected. However, the specific natures of the individual contributions of its various subregions to conscious memory processes remain equivocal. Here we show a functional dissociation between the hippocampus proper and the parahippocampal region in conscious and unconscious memory as revealed by invasive recordings of limbic event-related brain potentials recorded during explicit and implicit word recognition: Only hippocampal and not parahippocampal neural activity exhibits a sensitivity to the implicit versus explicit nature of the recognition memory task. Moreover, only within the hippocampus proper do the neural responses to repeated words differ not only from those to new words but also from each other as a function of recognition success. By contrast parahippocampal (rhinal) responses are sensitive to repetition independent of conscious recognition. These findings thus demonstrate that it is the hippocampus proper among the MTL structures that is specifically engaged during conscious memory processes.

Indications of nonlinear deterministic and finite-dimensional structures in time series of brain electrical activity: dependence on recording region and brain state.

We compare dynamical properties of brain electrical activity from different recording regions and from different physiological and pathological brain states. Using the nonlinear prediction error and an estimate of an effective correlation dimension in combination with the method of iterative amplitude adjusted surrogate data, we analyze sets of electroencephalographic (EEG) time series: surface EEG recordings from healthy volunteers with eyes closed and eyes open, and intracranial EEG recordings from epilepsy patients during the seizure free interval from within and from outside the seizure generating area as well as intracranial EEG recordings of epileptic seizures. As a preanalysis step an inclusion criterion of weak stationarity was applied. Surface EEG recordings with eyes open were compatible with the surrogates' null hypothesis of a Gaussian linear stochastic process. Strongest indications of nonlinear deterministic dynamics were found for seizure activity. Results of the other sets were found to be inbetween these two extremes.

Human memory formation is accompanied by rhinal-hippocampal coupling and decoupling.

In humans, distinct processes within the hippocampus and rhinal cortex support declarative memory formation. But do these medial temporal lobe (MTL) substructures directly cooperate in encoding new memories? Phase synchronization of gamma-band electroencephalogram (EEG) oscillations (around 40 Hz) is a general mechanism of transiently connecting neural assemblies. We recorded depth-EEG from within the MTL of epilepsy patients performing a memorization task. Successful as opposed to unsuccessful memory formation was accompanied by an initial elevation of rhinal-hippocampal gamma synchronization followed by a later desynchronization, suggesting that effective declarative memory formation is accompanied by a direct and temporarily limited cooperation between both MTL substructures.

Inferior temporal stream for word processing with integrated mnemonic function.

The participation of the inferior temporal cortex in visual word perception and recognition raises several questions: Is there a directed processing stream proceeding anteriorly by continuous cortical processing? How fast are words processed within such an inferior temporal stream? Does this stream support implicit or explicit memory? To answer these questions, we analyzed the spatio-temporal relationship of event-related potentials, recorded directly from the inferior temporal cortex in epilepsy patients performing a continuous visual word recognition paradigm. Event-related potentials elicited an inferior temporal positivity in a strip along the left collateral sulcus. This potential exhibited a linear (r = 0.74) peak latency progression from posterior to anterior inferior temporal regions (approximately 15 cm/sec), indicating a directed, intracortical processing stream. Peak amplitudes and latencies showed reliable old/new effects with smaller amplitudes and shorter latencies for old as opposed to new words. Although the amplitude-old/new-effect occurred for all repeated words (e.g., implicit memory), the latency-old/new-effect occurred for correctly recognized old words only (e.g., explicit recognition). These results seem to dissociate two distinct mnemonic processes. The graded decrease of mean ITP peak amplitudes and latencies, however, does not allow us to exclude a single trace model as assumed for explicit recognition memory based on familiarity (Mandler [1980]: Psychol Rev 87:252-271). Regardless whether there is a dissociation between implicit and explicit memory in inferior temporal cortex or not, our findings are in accordance with an integrated inferior temporal processing stream for words that performs continuously semantic and mnemonic operations supporting both implicit and explicit memory.

Nonlinear EEG analysis in epilepsy: its possible use for interictal focus localization, seizure anticipation, and prevention.

Several recent studies emphasize the high value of nonlinear EEG analysis particularly for improved characterization of epileptic brain states. In this review the authors report their work to increase insight into the spatial and temporal dynamics of the epileptogenic process. Specifically, they discuss possibilities for seizure anticipation, which is one of the most challenging aspects of epileptology. Although there are numerous studies exploring basic neuronal mechanisms that are likely to be associated with seizures, to date no definite information is available regarding how, when, or why a seizure occurs. Nonlinear EEG analysis now provides strong evidence that the interictal-ictal state transition is not an abrupt phenomenon. Rather, findings indicate that it is indeed possible to detect a preseizure phase. The unequivocal definition of such a state with a sufficient length would enable investigations of basic mechanisms leading to seizure initiation in humans, and development of adequate seizure prevention strategies.

Nonlinear noise reduction using reference data.

We introduce a method to clean uncorrelated deterministic and stochastic noise components from time series. It combines recently developed techniques for nonlinear projection with properties of the wavelet transform to extract noise in state space. The method requires that time series are generated by a dynamical system which is at least approximately deterministic and that they are recorded together with a reference signal. Its efficiency was tested on both simulated signals and measured magnetic fields of the heart. Convincing results are obtained even at low signal-to-noise ratios.

The epileptic process as nonlinear deterministic dynamics in a stochastic environment: an evaluation on mesial temporal lobe epilepsy.

The theory of deterministic chaos addresses simple deterministic dynamics in which nonlinearity gives rise to complex temporal behavior. Although biological neuronal networks such as the brain are highly complicated, a number of studies provide growing evidence that nonlinear time series analysis of brain electrical activity in patients with epilepsy is capable of providing potentially useful diagnostic information. In the present study, this analysis framework was extended by introducing a new measure xi, designed to discriminate between nonlinear deterministic and linear stochastic dynamics. For the evaluation of its discriminative power, xi was extracted from intracranial multi-channel EEGs recorded during the interictal state in 25 patients with unilateral mesial temporal lobe epilepsy. Strong indications of nonlinear determinism were found in recordings from within the epileptogenic zone, while EEG signals from other sites mainly resembled linear stochastic dynamics. In all investigated cases, this differentiation allowed to retrospectively determine the side of the epileptogenic zone in full agreement with results of the presurgical workup.

Time adaptive denoising of single trial event-related potentials in the wavelet domain.

We present a new wavelet-based method for single trial analysis of transient and time variant event-related potentials (ERPs). Expecting more accurate filter settings than achieved by other techniques (low-pass filter, a posteriori Wiener filter, time invariant wavelet filter), ERPs were initially balanced in time. By simulation, better filter performance could be established for test signals contaminated with either white noise or isospectral noise. To provide an example of real application, the method was applied to limbic P300 potentials (MTL-P300). As a result, variance of single trial MTL-P300s decreased, without restricting the corresponding mean. The proposed method can be regarded as an alternative for single-trial ERP analysis.

Single trial analysis of event related potentials: non-linear de-noising with wavelets.

We present and apply a method for single trial analysis of event related potentials (ERPs) that combines techniques from non-linear time series analysis with the wavelet transform. For this method, convincing results have already been achieved for simulated data as well as for intracranially recorded ERPs (Physica D 140 (2000) 257). However, ERPs are affected by a variety of external and internal experimental parameters, which makes the appropriate configuration of single trial analysis difficult. Thus, several pitfalls may occur in realistic applications. In this paper, we applied the method of non-linear de-noising with wavelets to both intra- and extracranially recorded ERPs, and show examples of how and where this single trial analysis can be used to obtain additional information on dynamic neural processes.

Nonlinear EEG analysis and its potential role in epileptology.

Deterministic chaos offers a striking explanation for apparently irregular behavior of the brain that is evidenced in the EEG. Recent developments in the physical-mathematical framework of the theory of nonlinear dynamics (colloquially often termed chaos theory) provide new concepts and powerful algorithms to analyze such time series. Because of its high versatility, nonlinear time series analysis has already gone beyond the physical sciences and, at present, is being successfully applied in a variety of disciplines, including cardiology, neurology, psychiatry, and epileptology. However, it is well known that different influencing factors limit the use of nonlinear measures to characterize EEG dynamics in a strict sense. Nevertheless, when interpreted with care, relative estimates of, e.g., the correlation dimension or the Lyapunov exponents, can reliably characterize different states of normal and pathologic brain function. In epileptology, extraction of nonlinear measures from the intracranially recorded EEG promises to be important for clinical practice. In addition to an immense reduction of information content of long-lasting EEG recordings, previous studies have shown that these measures enable (a) localization of the primary epileptogenic area in different cerebral regions during the interictal state, (b) investigations of antiepileptic drug effects, (c) analyses of spatio-temporal interactions between the epileptogenic zone and other brain areas, and (d) detection of features predictive of imminent seizure activity. Nonlinear time series analysis provides new and supplementary information about the epileptogenic process and thus contributes to an improvement in presurgical evaluation.

Spatial distribution of neuronal complexity loss in neocortical lesional epilepsies.

PURPOSE: Nonlinear EEG analysis is valuable in characterizing the spatiotemporal dynamics of the epileptogenic process in mesial temporal lobe epilepsy. We examined the ability of the measure neuronal complexity loss (L*) to characterize the primary epileptogenic area of neocortical lesional epilepsies during the interictal state. METHODS: Spatial distribution of L* (L* map) was extracted from electrocorticograms (n = 52) recorded during presurgical assessment via subdural 64-contact grid electrodes covering lesions in either frontal, parietal, or temporal neocortex in 15 patients. The exact location of recording contacts on the brain surface was identified by matching a postimplant lateral x-ray of the skull with a postoperatively obtained sagittal MRI scan. Reprojecting L* maps onto the subject's brain surface allowed us to compare the spatial distribution of L* with the resection range of the extended lesionectomy. RESULTS: In each of the six patients who became seizure-free, maximum values of L* were restricted to recording sites coinciding with the area of resection. In contrast, L* maps of most patients who had no benefit from the resection indicated a more widespread extent or the existence of additional, probably autonomous, foci. The mean of L* values obtained from recording sites outside the area of resection correctly distinguished 13 patients (86.7 %) with respect to seizure outcome. CONCLUSIONS: Relevant information obtained from long-lasting interictal electrocorticographic recordings can be compressed to a single L* map that contributes to a spatial characterization of the primary epileptogenic area. In neocortical lesional epilepsies, L* allows for identification and characterization of epileptogenic activity and thus provides an additional diagnostic tool for presurgical assessment.

Kulback-Leibler and renormalized entropies: applications to electroencephalograms of epilepsy patients.

Recently, "renormalized entropy" was proposed as a novel measure of relative entropy [P. Saparin et al., Chaos, Solitons and Fractals 4, 1907 (1994)] and applied to several physiological time sequences, including electroencephalograms (EEGs) of patients with epilepsy. We show here that this measure is just a modified Kullback-Leibler (KL) relative entropy, and it gives similar numerical results to the standard KL entropy. The latter better distinguishes frequency contents of, e.g., seizure and background EEGs than renormalized entropy. We thus propose that renormalized entropy might not be as useful as claimed by its proponents. In passing, we also make some critical remarks about the implementation of these methods.

Non-linear time series analysis of intracranial EEG recordings in patients with epilepsy--an overview.

Deterministic chaos offers a striking explanation for apparently irregular behavior, a characteristic feature of brain electrical activity. The framework of the theory of non-linear dynamics provides new concepts and powerful algorithms to analyze such time series. However, different influencing factors render the use of non-linear measures in a strict sense problematic. Nevertheless, if interpreted with care, particularly the correlation dimension or the Lyapunov-exponents provide a means to reliably characterize different states of normal and pathological brain function. This overview summarizes recent findings applying this concept in the field of epileptology that promise to be important for clinical practice. Non-linear measures extracted from the intra-cranially recorded EEG allow (a) localization of epileptogenic areas in different cerebral regions even during seizure-free intervals, (b) investigation of the influence of anticonvulsive drugs and (c) detection of features predictive of imminent seizure activity. Moreover, particularly the dimensional complexity proves a valuable parameter reflecting spatially distributed neuronal activity during verbal learning and memory processes. Specific changes in time of this non-linear measure allow the prediction of memory performance and, in addition, represent an estimate of the recruitment potency in the anterior mesial temporal lobes. Thus, the application of non-linear time series analysis to brain electrical activity offers new information about the dynamics of the underlying neuronal networks.

Evidence relating human verbal memory to hippocampal N-methyl-D-aspartate receptors.

Studies in rodents and nonhuman primates have linked the activity of N-methyl-D-aspartate (NMDA) receptors within the hippocampus to animals' performance on memory-related tasks. However, whether these receptors are similarly essential for human memory is still an open question. Here we present evidence suggesting that hippocampal NMDA receptors, most likely within the CA1 region, do participate in human verbal memory processes. Words elicit a negative event-related potential (ERP) peaking around 400 ms within the anterior mesial temporal lobe (AMTL-N400). Ketamine, an NMDA-receptor antagonist, reduces the amplitude of the AMTL-N400 (in contrast to other hippocampal potentials) on initial presentation, eliminates the typical AMTL-N400 amplitude reduction with repetition, and leads to significant memory impairment. Of the various hippocampal subfields, only the density of CA1 neurons correlates with the word-related ERPs that are reduced by ketamine. Altogether, our behavioral, anatomical, and electrophysiological results indicate that hippocampal NMDA receptors are involved in human memory.

Is it possible to anticipate seizure onset by non-linear analysis of intracerebral EEG in human partial epilepsies?

Detection of electrophysiological features preceding and indicative of imminent seizures in patients with epilepsy would be a major breakthrough with immense scientific and clinical implications. The definition of a "pre-ictal state" several minutes prior to seizure onset would open a new time window for studying mechanisms of seizure generation as well as for possible therapeutic interventions. In this review we present recent findings from nonlinear time series analysis of intracranially recorded EEG that may allow to forecast seizure onset in patients with partial epilepsy.

Real-time tracking of memory formation in the human rhinal cortex and hippocampus.

A fundamental question about human memory is which brain structures are involved, and when, in transforming experiences into memories. This experiment sought to identify neural correlates of memory formation with the use of intracerebral electrodes implanted in the brains of patients with temporal lobe epilepsy. Event-related potentials (ERPs) were recorded directly from the medial temporal lobe (MTL) as the patients studied single words. ERPs elicited by words subsequently recalled in a memory test were contrasted with ERPs elicited by unrecalled words. Memory formation was associated with distinct but interrelated ERP differences within the rhinal cortex and the hippocampus, which arose after about 300 and 500 milliseconds, respectively. These findings suggest that declarative memory formation is dissociable into subprocesses and sequentially organized within the MTL.

Reduced signal complexity of intracellular recordings: a precursor for epileptiform activity?

Recent studies have shown that time windowed extraction of nonlinear parameters like an effective correlation dimension from intracranially recorded EEG of epileptic patients often allows to detect and identify an unequivocal "pre-ictal phase" preceding an epileptic seizure. In another study, however, such an anticipation could not be made. These conflicting findings may indicate that observed changes in nonlinear parameters probably depend on the type of elementary mechanisms underlying epileptic processes and/or the spatial distribution of neurons primarily involved in generation of epileptiform discharges. To test the existence of such dependencies, the transition from normal to epileptiform activity (EA) of CA3-neurons in hippocampal slices was analyzed in four epilepsy models, using a time windowed computation of an effective correlation dimension. Indeed, in xanthine and penicillin models, signal complexity in intracellular recordings was reduced before manifestation of paroxysmal depolarization shifts (PDS), whereas a preceding loss of complexity was missing in low-magnesium and veratridine models. These findings indicate that interictal-like EA is predictable only in some epilepsy models.