Spindler: a framework for parametric analysis and detection of spindles in EEG with application to sleep spindles.

OBJECTIVE: EEG spindles, narrow-band oscillatory signal bursts, are widely-studied biomarkers of subject state and neurological function. Most existing methods for spindle detection select algorithm parameters by optimizing agreement with expert labels. We propose a new framework for selecting algorithm parameters based on stability of spindle properties and elucidate the dependence of these properties on parameter selection for several algorithms. APPROACH: To demonstrate this approach we developed a new algorithm (Spindler) that decomposes the signal using matching pursuit with Gabor atoms and computes the spindles for each point in a fine grid of parameter values. After computing characteristic surfaces as a function of parameters, Spindler selects algorithm parameters based on the stability of characteristic surface geometry. MAIN RESULTS: Spindler performs well relative to several common supervised and unsupervised EEG sleep spindle detection methods. Spindler is available as an open-source MATLAB toolbox (https://github.com/VisLab/EEG-Spindles). In addition to Spindler, the toolbox provides implementations of several other spindle detection algorithms as well as standardized methods for matching ground truth to predictions and a framework for understanding algorithm parameter surfaces. SIGNIFICANCE: This work demonstrates that parameter selection based on physical constraints rather than labelled data can provide effective, fully-automated, unsupervised spindle detection. This work also exposes the dangers of applying cross-validation without considering the dependence of spindle properties on parameters. Parameters selected to optimize one performance metric or matching method are not optimized for others. Furthermore, elucidation of the stability of predicted indicators with respect to algorithm parameter selection is critical to practical application of these algorithms.

Ictal propagation of high frequency activity is recapitulated in interictal recordings: effective connectivity of epileptogenic networks recorded with intracranial EEG.

Seizures are increasingly understood to arise from epileptogenic networks across which ictal activity is propagated and sustained. In patients undergoing invasive monitoring for epilepsy surgery, high frequency oscillations have been observed within the seizure onset zone during both ictal and interictal intervals. We hypothesized that the patterns by which high frequency activity is propagated would help elucidate epileptogenic networks and thereby identify network nodes relevant for surgical planning. Intracranial EEG recordings were analyzed with a multivariate autoregressive modeling technique (short-time direct directed transfer function--SdDTF), based on the concept of Granger causality, to estimate the directionality and intensity of propagation of high frequency activity (70-175 Hz) during ictal and interictal recordings. These analyses revealed prominent divergence and convergence of high frequency activity propagation at sites identified by epileptologists as part of the ictal onset zone. In contrast, relatively little propagation of this activity was observed among the other analyzed sites. This pattern was observed in both subdural and depth electrode recordings of patients with focal ictal onset, but not in patients with a widely distributed ictal onset. In patients with focal ictal onsets, the patterns of propagation recorded during pre-ictal (up to 5 min immediately preceding ictal onset) and interictal (more than 24h before and after seizures) intervals were very similar to those recorded during seizures. The ability to characterize epileptogenic networks from interictal recordings could have important clinical implications for epilepsy surgery planning by reducing the need for prolonged invasive monitoring to record spontaneous seizures.

Applying EEG phase synchronization measures to non-linearly coupled neural mass models.

BACKGROUND: Recent neuroimaging analyses aim to understand how information is integrated across brain regions that have traditionally been studied in isolation; however, detecting functional connectivity networks in experimental EEG recordings is a non-trivial task. NEW METHOD: We use neural mass models to simulate 10-s trials with coupling between 1-3 and 5-8s and compare how well three phase-based connectivity measures recover this connectivity pattern across a set of experimentally relevant conditions: variable oscillation frequency and power spectrum, feed forward connections with or without feedback, and simulated signals with and without volume conduction. RESULTS: Overall, the results highlight successful detection of the onset and offset of significant synchronizations for a majority of the 28 simulated configurations; however, the tested phase measures sometimes differ in their sensitivity and specificity to the underlying connectivity. COMPARISON WITH EXISTING METHODS: Prior work has shown that these phase measures perform well on signals generated by a computational model of coupled oscillators. In this work we extend previous studies by exploring the performance of these measures on a different class of computational models, and we compare the methods on 28 variations that capture a set of experimentally relevant conditions. CONCLUSIONS: Our results underscore that no single phase synchronization measure is substantially better than all others, and experimental investigations will likely benefit from combining a set of measures together that are chosen based on both the experimental question of interest, the signal to noise ratio in the EEG data, and the approach used for statistical significance.

Comparing parametric and nonparametric methods for detecting phase synchronization in EEG.

Detecting significant periods of phase synchronization in EEG recordings is a non-trivial task that is made especially difficult when considering the effects of volume conduction and common sources. In addition, EEG signals are often confounded by non-neural signals, such as artifacts arising from muscle activity or external electrical devices. A variety of phase synchronization analysis methods have been developed with each offering a different approach for dealing with these confounds. We investigate the use of a parametric estimation of the time-frequency transform as a means of improving the detection capability for a range of phase analysis methods. We argue that such an approach offers numerous benefits over using standard nonparametric approaches. We then demonstrate the utility of our technique using both simulated and actual EEG data by showing that the derived phase synchronization estimates are more robust to noise and volume conduction effects.

Fear conditioning is associated with dynamic directed functional interactions between and within the human amygdala, hippocampus, and frontal lobe.

The current model of fear conditioning suggests that it is mediated through modules involving the amygdala (AMY), hippocampus (HIP), and frontal lobe (FL). We now test the hypothesis that habituation and acquisition stages of a fear conditioning protocol are characterized by different event-related causal interactions (ERCs) within and between these modules. The protocol used the painful cutaneous laser as the unconditioned stimulus and ERC was estimated by analysis of local field potentials recorded through electrodes implanted for investigation of epilepsy. During the prestimulus interval of the habituation stage FL>AMY ERC interactions were common. For comparison, in the poststimulus interval of the habituation stage, only a subdivision of the FL (dorsolateral prefrontal cortex, dlPFC) still exerted the FL>AMY ERC interaction (dlFC>AMY). For a further comparison, during the poststimulus interval of the acquisition stage, the dlPFC>AMY interaction persisted and an AMY>FL interaction appeared. In addition to these ERC interactions between modules, the results also show ERC interactions within modules. During the poststimulus interval, HIP>HIP ERC interactions were more common during acquisition, and deep hippocampal contacts exerted causal interactions on superficial contacts, possibly explained by connectivity between the perihippocampal gyrus and the HIP. During the prestimulus interval of the habituation stage, AMY>AMY ERC interactions were commonly found, while interactions between the deep and superficial AMY (indirect pathway) were independent of intervals and stages. These results suggest that the network subserving fear includes distributed or widespread modules, some of which are themselves "local networks." ERC interactions between and within modules can be either static or change dynamically across intervals or stages of fear conditioning.

Painful laser stimuli induce directed functional interactions within and between the human amygdala and hippocampus.

The pathways by which painful stimuli are signaled within the human medial temporal lobe are unknown. Rodent studies have shown that nociceptive inputs are transmitted from the brainstem or thalamus through one of two pathways to the central nucleus of the amygdala. The indirect pathway projects from the basal and lateral nuclei of the amygdala to the central nucleus, while the direct pathway projects directly to the central nucleus. We now test the hypothesis that the human ventral amygdala (putative basal and lateral nuclei) exerts a causal influence upon the dorsal amygdala (putative central nucleus), during the application of a painful laser stimulus. Local field potentials (LFPs) were recorded from depth electrode contacts implanted in the medial temporal lobe for the treatment of epilepsy, and causal influences were analyzed by Granger causality (GRC). This analysis indicates that the dorsal amygdala exerts a pre-stimulus causal influence upon the hippocampus, consistent with an attention-related response to the painful laser. Within the amygdala, the analysis indicates that the ventral contacts exert a causal influence upon dorsal contacts, consistent with the human (putative) indirect pathway. Potentials evoked by the laser (LEPs) were not recorded in the ventral nuclei, but were recorded at dorsal amygdala contacts which were not preferentially those receiving causal influences from the ventral contacts. Therefore, it seems likely that the putative indirect pathway is associated with causal influences from the ventral to the dorsal amygdala, and is distinct from the human (putative) indirect pathway which mediates LEPs in the dorsal amygdala.

Attention to painful cutaneous laser stimuli evokes directed functional connectivity between activity recorded directly from human pain-related cortical structures.

Our previous studies show that attention to painful cutaneous laser stimuli is associated with functional connectivity between human primary somatosensory cortex (SI), parasylvian cortex (PS), and medial frontal cortex (MF), which may constitute a pain network. However, the direction of functional connections within this network is unknown. We now test the hypothesis that activity recorded from the SI has a driver role, and a causal influence, with respect to activity recorded from PS and MF during attention to a laser. Local field potentials (LFP) were recorded from subdural grid electrodes implanted for the treatment of epilepsy. We estimated causal influences by using the Granger causality (GRC), which was computed while subjects performed either an attention task (counting laser stimuli) or a distraction task (reading for comprehension). Before the laser stimuli, directed attention to the painful stimulus (counting) consistently increased the number of GRC pairs both within the SI cortex and from SI upon PS (SI>PS). After the laser stimulus, attention to a painful stimulus increased the number of GRC pairs from SI>PS, and SI>MF, and within the SI area. LFP at some electrode sites (critical sites) exerted GRC influences upon signals at multiple widespread electrodes, both in other cortical areas and within the area where the critical site was located. Critical sites may bind these areas together into a pain network, and disruption of that network by stimulation at critical sites might be used to treat pain. Electrical activity recorded from the somatosensory cortex drives activity recorded elsewhere in the pain network and may bind the network together; disruption of that network by stimulation at critical sites might be used to treat pain.

Intracranial mapping of auditory perception: event-related responses and electrocortical stimulation.

OBJECTIVE: We compared intracranial recordings of auditory event-related responses with electrocortical stimulation mapping (ESM) to determine their functional relationship. METHODS: Intracranial recordings and ESM were performed, using speech and tones, in adult epilepsy patients with subdural electrodes implanted over lateral left cortex. Evoked N1 responses and induced spectral power changes were obtained by trial averaging and time-frequency analysis. RESULTS: ESM impaired perception and comprehension of speech, not tones, at electrode sites in the posterior temporal lobe. There was high spatial concordance between ESM sites critical for speech perception and the largest spectral power (100% concordance) and N1 (83%) responses to speech. N1 responses showed good sensitivity (0.75) and specificity (0.82), but poor positive predictive value (0.32). Conversely, increased high-frequency power (>60Hz) showed high specificity (0.98), but poorer sensitivity (0.67) and positive predictive value (0.67). Stimulus-related differences were observed in the spatial-temporal patterns of event-related responses. CONCLUSIONS: Intracranial auditory event-related responses to speech were associated with cortical sites critical for auditory perception and comprehension of speech. SIGNIFICANCE: These results suggest that the distribution and magnitude of intracranial auditory event-related responses to speech reflect the functional significance of the underlying cortical regions and may be useful for pre-surgical functional mapping.

Estimation of short-time cross-correlation between frequency bands of event related EEG.

Simultaneous variations of the event-related power changes (ERD/ERS) are often observed in a number of frequency bands. ERD/ERS measures are usually based on the relative changes of power in a given single frequency band. Within such an approach one cannot answer questions concerning the mutual relations between the band-power variations observed in different frequency bands. This paper addresses the problem of estimating and assessing the significance of the average cross-correlation between ERD/ERS phenomena occurring in two frequency bands. The cross-correlation function in a natural way also provides estimation of the delay between ERD/ERS in those bands. The proposed method is based on estimating the short-time cross-correlation function between relative changes of power in two selected frequency bands. The cross-correlation function is estimated in each trial separately and then averaged across trials. The significance of those mean cross-correlation functions is evaluated by means of a nonparametric test. The basic properties of the method are presented on simulated signals, and an example application to real EEG and ECoG signals is given.

Computationally efficient approaches to calculating significant ERD/ERS changes in the time-frequency plane.

This paper addresses some practical issues related to the calculation, display and assessment of the significance of changes in the average time-frequency energy density of event-related brain activity. Using scalp EEG and subdural ECoG example datasets, parametric tests are evaluated as a replacement for previously applied computer-intensive resampling methods. The performance of different estimates of energy density, based on matching pursuit, scalogram and spectrogram, and their Box-Cox transformations is evaluated with respect to the assumption of normality required for the t-test, and the consistency of the final results.

Gabor atom density as a measure of seizure complexity.

The Gabor atom density (GAD) is a measure of complexity of a signal. It is based on the time-frequency decomposition obtained by the matching pursuit (MP) algorithm. The GAD/MP method was applied to EEG data recorded from intracranial electrodes in patients with intractable complex partial seizures. GAD shows that epileptic seizures, which are reflections of increased neuronal synchrony, are also periods of increased and changing signal complexity. The GAD/MP method is well suited to analyzing these signals from seizures characterized by rapid dynamical changes. The period of organized rhythmic activity exhibits lower complexity than that seen during other phases of the seizure.

Software system for data management and distributed processing of multichannel biomedical signals.

The presented software is designed for efficient utilization of cluster of PC computers for signal analysis of multichannel physiological data. The system consists of three main components: 1) a library of input and output procedures, 2) a database storing additional information about location in a storage system, 3) a user interface for selecting data for analysis, choosing programs for analysis, and distributing computing and output data on cluster nodes. The system allows for processing multichannel time series data in multiple binary formats. The description of data format, channels and time of recording are included in separate text files. Definition and selection of multiple channel montages is possible. Epochs for analysis can be selected both manually and automatically. Implementation of a new signal processing procedures is possible with a minimal programming overhead for the input/output processing and user interface. The number of nodes in cluster used for computations and amount of storage can be changed with no major modification to software. Current implementations include the time-frequency analysis of multiday, multichannel recordings of intracranial EEG of epileptic patients as well as evoked response analyses of repeated cognitive tasks.

Epileptic seizures are characterized by changing signal complexity.

OBJECTIVE: Epileptic seizures are brief episodic events resulting from abnormal synchronous discharges from cerebral neuronal networks. The traditional methods of signal analysis are limited by the rapidly changing nature of the EEG signal during a seizure. Time-frequency analyses, however, such as those produced by the matching pursuit (MP) method can provide continuous decompositions of recorded seizure activity. These accurate decompositions can allow for more detailed analyses of the changes in complexity of the signal that may accompany seizure evolution. METHODS: The MP algorithm was applied to provide time-frequency decompositions of entire seizures recorded from depth electrode contacts in patients with intractable complex partial seizures of mesial temporal onset. The results of these analyses were compared with signals generated from the Duffing equation that represented both limit cycle and chaotic behavior. RESULTS: Seventeen seizures from 12 different patients were analyzed. These analyses reveal that early in the seizure, the most organized, rhythmic seizure activity is more complex than limit cycle behavior, and that signal complexity increases further later in the seizure. CONCLUSIONS: Increasing complexity routinely precedes seizure termination. This may reflect progressive desynchronization.

An autoregressive method for the measurement of synchronization of interictal and ictal EEG signals.

We propose a new measure of synchronization of multichannel ictal and interictal EEG signals. The measure is based on the residual covariance matrix of a multichannel autoregressive model. A major advantage of this measure is its ability to be interpreted both in the framework of stochastic and deterministic models. A preliminary analysis of EEG data from three patients using this measure documents the expected increased synchronization during ictal periods but also reveals that increased synchrony persists for prolonged periods (up to 2 h or more) in the postictal period.

Application of the directed transfer function method to mesial and lateral onset temporal lobe seizures.

The directed transfer function (DTF) method is a multichannel analysis based on an autoregressive model that detects flow of seizure activity. This report extends the application of the DTF method to compare patterns of flow of seizures with different sites of origin. Analysis of a seizure originating from mesial temporal structures is compared with a seizure originating from lateral temporal neocortex; both complex partial seizures were recorded with intracranial electrodes that combine subdural grid arrays and depth electrodes. The DTF method has the potential to determine patterns of flow of activity, including periods when visual analysis of the intracranial ictal EEG may not allow for definitive source localization. The extension of the DTF analyses into integrated DTF (IDTF) formats is also illustrated. When activity of a relatively discrete frequency can be identified, the IDTF analysis facilitates display of patterns of flow of this selected activity.

Time-frequency analysis using the matching pursuit algorithm applied to seizures originating from the mesial temporal lobe.

OBJECTIVES: The ability to analyze patterns of recorded seizure activity is important in the localization and classification of seizures. Ictal evolution is typically a dynamic process with signals composed of multiple frequencies; this can limit or complicate methods of analysis. The recently-developed matching pursuit algorithm permits continuous time-frequency analyses, making it particularly appealing for application to these signals. The studies here represent the initial applications of this method to intracranial ictal recordings. METHODS: Mesial temporal onset partial seizures were recorded from 9 patients. The data were analyzed by the matching pursuit algorithm were continuous digitized single channel recordings from the depth electrode contact nearest the region of seizure onset. Tine frequency energy distributions were plotted for each seizure and correlated with the intracranial EEG recordings. RESULTS: Periods of seizure initiation, transitional rhythmic bursting activity, organized rhythmic bursting activity and intermittent bursting activity were identified. During periods of organized rhythmic bursting activity, all mesial temporal onset seizures analyzed had a maximum predominant frequency of 5.3-8.4 Hz with a monotonic decline in frequency over a period of less than 60 s. The matching pursuit method allowed for time-frequency decomposition of entire seizures. CONCLUSIONS: The matching pursuit method is a valuable tool for time-frequency analyses of dynamic seizure activity. It is well suited for application to the non-stationary activity that typically characterizes seizure evolution. Time-frequency patterns of seizures originating from different brain regions can be compared using the matching pursuit method.

A simple computer model of excitable synaptically connected neurons.

The space-lumped two-variable neuron model is studied. Extension of the neural model by adding a simple synaptic current allows the demonstration of neural interactions. The production of synchronous burst activity in this simple two-neuron excitatory loop is modeled, including the influence of random background excitatory input. The ability of the neuron model to integrate inputs spatially and temporally is shown. Two refractory periods after stimuli were identified and their role in burst cessation is demonstrated. Our findings show that simple neural units without long-lasting membrane processes are capable of generating long lasting patterns of activity. The results of simulation of simple background activity suggest that an increase in background activity tends to cause decreased activity of the network. This phenomenon, as well as the existence of two refractory periods, allows for burst cessation without inhibition in this simple model.

A general statistical framework for frequency-domain analysis of EEG topographic structure.

A wide variety of rhythmic electrophysiological phenomena--including driven, induced, and endogenous activities of cortical neuronal masses--lend themselves naturally to analysis using frequency--domain techniques applied to multichannel recordings that discretely sample the overall spatial pattern of the rhythmic activity. For such cases, a large but so far poorly utilized body of statistical theory supports a third major approach to topographic analysis, complementing the more familiar mapping and source-recovery techniques. These methods, many of which have only recently become computationally feasible, collectively provide general solutions to the problem of detecting and characterizing systematic differences that arise--not only in the spatial distribution of the activity, but also in its frequency-dependent between-channel covariance structure--as a function of multiple experimental conditions presented in conformity with any of the conventional experimental designs. This application-oriented tutorial review provides a comprehensive outline of these resources, including: (1) real multivariate analysis of single-channel spectral measures (and measures of between-channel relationships such as coherence and phase), (2) complex multivariate analysis based on multichannel Fourier transforms, and (3) complex multivariate analysis based on multichannel parametric models. Special emphasis is placed on the potential of the multichannel autoregressive model to support EEG (and MEG) studies of perceptual and cognitive processes.

Detection of spontaneous postsynaptic potentials.

The amplitude and time course of postsynaptic potentials (PSPs) recorded by intracellular techniques contain information that allow different synaptic events to be detected. In the present paper an algorithm to detect spontaneous PSPs is described. The algorithm is based on computation of approximations of first and second derivatives of the signals. The method was tested on both computer-simulated potentials and on experimental data recorded from dissociated mouse spinal cord neurons in tissue culture. The receiver operating characteristics of the detection algorithm were computed. This method can be applied to investigations of dynamic changes in the activity of neural networks.

Analysis of mesial temporal seizure onset and propagation using the directed transfer function method.

The directed transfer function (DTF) method, a multichannel parametric method of analysis based on an autoregressive model, is a newly developed tool that permits determination of patterns of flow of activity. The DTF method of analysis was applied to seizures originating from mesial temporal lobe structures in 3 patients recorded by combined subdural grid and depth electrode arrays. These first applications to human intracranial recordings demonstrated that the DTF method can accurately determine patterns of seizure onset and propagation. In addition the DTF method can provide evidence regarding patterns of flow of seizure activity that are not readily apparent from visual inspection of the EEG recordings. Important considerations for appropriate application of the DTF method for the analysis of intracranial ictal recordings are discussed.