Optimization of data pre-processing methods for time-series classification of electroencephalography data.

The performance of time-series classification of electroencephalographic data varies strongly across experimental paradigms and study participants. Reasons are task-dependent differences in neuronal processing and seemingly random variations between subjects, amongst others. The effect of data pre-processing techniques to ameliorate these challenges is relatively little studied. Here, the influence of spatial filter optimization methods and non-linear data transformation on time-series classification performance is analyzed by the example of high-frequency somatosensory evoked responses. This is a model paradigm for the analysis of high-frequency electroencephalography data at a very low signal-to-noise ratio, which emphasizes the differences of the explored methods. For the utilized data, it was found that the individual signal-to-noise ratio explained up to 74% of the performance differences between subjects. While data pre-processing was shown to increase average time-series classification performance, it could not fully compensate the signal-to-noise ratio differences between the subjects. This study proposes an algorithm to prototype and benchmark pre-processing pipelines for a paradigm and data set at hand. Extreme learning machines, Random Forest, and Logistic Regression can be used quickly to compare a set of potentially suitable pipelines. For subsequent classification, however, machine learning models were shown to provide better accuracy.

Sex-specific reference values for total, central, and peripheral latency of motor evoked potentials from a large cohort.

Background: Differentiating between physiologic and altered motor evoked potentials (MEPs) to transcranial magnetic stimulation (TMS) is crucial in clinical practice. Some physical characteristics, such as height and age, introduce sources of variability unrelated to neural dysfunction. We provided new age- and height-adjusted normal values for cortical latency, central motor conduction time (CMCT), and peripheral motor conduction time (PMCT) from a large cohort of healthy subjects. Methods: Previously reported data from 587 participants were re-analyzed. Nervous system disorders were ruled out by clinical examination and magnetic resonance imaging. MEP latency was determined as stimulus-to-response latency through stimulation with a circular coil over the "hot spot" of the First Dorsal Interosseous and Tibialis Anterior muscles, during mild tonic contraction. CMCT was estimated as the difference between MEP cortical latency and PMCT by radicular magnetic stimulation. Additionally, right-to-left differences were calculated. For each parameter, multiple linear regression models of increasing complexity were fitted using height, age, and sex as regressors. Results: Motor evoked potential cortical latency, PMCT, and CMCT were shown to be age- and height-dependent, although age had only a small effect on CMCT. Relying on Bayesian information criterion for model selection, MEP cortical latency and PMCT were explained best by linear models indicating a positive correlation with both height and age. Also, CMCT to lower limbs positively correlated with height and age. CMCT to upper limbs positively correlated to height, but slightly inversely correlated to age, as supported by non-parametric bootstrap analysis. Males had longer cortical latencies and CMCT to lower limbs, as well as longer PMCT and cortical latencies to upper limbs, even when accounting for differences in body height. Right-to-left-differences were independent of height, age, and sex. Based on the selected regression models, sex-specific reference values were obtained for all TMS-related latencies and inter-side differences, with adjustments for height and age, where warranted. Conclusion: A significant relationship was observed between height and age and all MEP latency values, in both upper and lower limbs. These set of reference values facilitate the evaluation of MEPs in clinical studies and research settings. Unlike previous reports, we also highlighted the contribution of sex.

Separating Neural Oscillations from Aperiodic 1/f Activity: Challenges and Recommendations.

Electrophysiological power spectra typically consist of two components: An aperiodic part usually following an 1/f power law [Formula: see text] and periodic components appearing as spectral peaks. While the investigation of the periodic parts, commonly referred to as neural oscillations, has received considerable attention, the study of the aperiodic part has only recently gained more interest. The periodic part is usually quantified by center frequencies, powers, and bandwidths, while the aperiodic part is parameterized by the y-intercept and the 1/f exponent [Formula: see text]. For investigation of either part, however, it is essential to separate the two components. In this article, we scrutinize two frequently used methods, FOOOF (Fitting Oscillations & One-Over-F) and IRASA (Irregular Resampling Auto-Spectral Analysis), that are commonly used to separate the periodic from the aperiodic component. We evaluate these methods using diverse spectra obtained with electroencephalography (EEG), magnetoencephalography (MEG), and local field potential (LFP) recordings relating to three independent research datasets. Each method and each dataset poses distinct challenges for the extraction of both spectral parts. The specific spectral features hindering the periodic and aperiodic separation are highlighted by simulations of power spectra emphasizing these features. Through comparison with the simulation parameters defined a priori, the parameterization error of each method is quantified. Based on the real and simulated power spectra, we evaluate the advantages of both methods, discuss common challenges, note which spectral features impede the separation, assess the computational costs, and propose recommendations on how to use them.

Electrophysiological characterization of the hyperdirect pathway and its functional relevance for subthalamic deep brain stimulation.

The subthalamic nucleus (STN) receives input from various cortical areas via hyperdirect pathway (HDP) which bypasses the basal-ganglia loop. Recently, the HDP has gained increasing interest, because of its relevance for STN deep brain stimulation (DBS). To understand the HDP's role cortical responses evoked by STN-DBS have been investigated. These responses have short (<2 ms), medium (2-15 ms), and long (20-70 ms) latencies. Medium-latency responses are supposed to represent antidromic cortical activations via HDP. Together with long-latency responses the medium responses can potentially be used as biomarker of DBS efficacy as well as side effects. We here propose that the activation sequence of the cortical evoked responses can be conceptualized as high frequency oscillations (HFO) for signal analysis. HFO might therefore serve as marker for antidromic activation. Using existing knowledge on HFO recordings, this approach allows data analyses and physiological modeling to advance the pathophysiological understanding of cortical DBS-evoked high-frequency activity.

Noninvasive neuromagnetic single-trial analysis of human neocortical population spikes.

Neuronal spiking is commonly recorded by invasive sharp microelectrodes, whereas standard noninvasive macroapproaches (e.g., electroencephalography [EEG] and magnetoencephalography [MEG]) predominantly represent mass postsynaptic potentials. A notable exception are low-amplitude high-frequency (∼600 Hz) somatosensory EEG/MEG responses that can represent population spikes when averaged over hundreds of trials to raise the signal-to-noise ratio. Here, a recent leap in MEG technology-featuring a factor 10 reduction in white noise level compared with standard systems-is leveraged to establish an effective single-trial portrayal of evoked cortical population spike bursts in healthy human subjects. This time-resolved approach proved instrumental in revealing a significant trial-to-trial variability of burst amplitudes as well as time-correlated (∼10 s) fluctuations of burst response latencies. Thus, ultralow-noise MEG enables noninvasive single-trial analyses of human cortical population spikes concurrent with low-frequency mass postsynaptic activity and thereby could comprehensively characterize cortical processing, potentially also in diseases not amenable to invasive microelectrode recordings.

Temporal Signatures of Criticality in Human Cortical Excitability as Probed by Early Somatosensory Responses.

Brain responses vary considerably from moment to moment, even to identical sensory stimuli. This has been attributed to changes in instantaneous neuronal states determining the system's excitability. Yet the spatiotemporal organization of these dynamics remains poorly understood. Here we test whether variability in stimulus-evoked activity can be interpreted within the framework of criticality, which postulates dynamics of neural systems to be tuned toward the phase transition between stability and instability as is reflected in scale-free fluctuations in spontaneous neural activity. Using a novel noninvasive approach in 33 male human participants, we tracked instantaneous cortical excitability by inferring the magnitude of excitatory postsynaptic currents from the N20 component of the somatosensory evoked potential. Fluctuations of cortical excitability demonstrated long-range temporal dependencies decaying according to a power law across trials, a hallmark of systems at critical states. As these dynamics covaried with changes in prestimulus oscillatory activity in the alpha band (8-13 Hz), we establish a mechanistic link between ongoing and evoked activity through cortical excitability and argue that the co-emergence of common temporal power laws may indeed originate from neural networks poised close to a critical state. In contrast, no signatures of criticality were found in subcortical or peripheral nerve activity. Thus, criticality may represent a parsimonious organizing principle of variability in stimulus-related brain processes on a cortical level, possibly reflecting a delicate equilibrium between robustness and flexibility of neural responses to external stimuli.SIGNIFICANCE STATEMENT Variability of neural responses in primary sensory areas is puzzling, as it is detrimental to the exact mapping between stimulus features and neural activity. However, such variability can be beneficial for information processing in neural networks if it is of a specific nature, namely, if dynamics are poised at a so-called critical state characterized by a scale-free spatiotemporal structure. Here, we demonstrate the existence of a link between signatures of criticality in ongoing and evoked activity through cortical excitability, which fills the long-standing gap between two major directions of research on neural variability: the impact of instantaneous brain states on stimulus processing on the one hand and the scale-free organization of spatiotemporal network dynamics of spontaneous activity on the other.

Cortical somatosensory evoked high-frequency (600Hz) oscillations predict absence of severe hypoxic encephalopathy after resuscitation.

OBJECTIVE: Following cardiac arrest (CA), hypoxic encephalopathy (HE) frequently occurs and hence reliable neuroprognostication is crucial to decide on the extent of intensive care. Several investigations predict severe HE leading to persistent unresponsive wakefulness or death, with high specificity. Only few studies attempted to predict absence of severe HE. Cortical somatosensory evoked high-frequency (600Hz) oscillation (HFO) bursts indicate the presence of highly synchronized spiking activity in the primary somatosensory cortex. Since global neuronal damage characterizes severe HE preserved cortical HFOs may early exclude severe HE. METHODS: We determined amplitudes of early and late HFO bursts in 302 comatose CA patients after median nerve somatosensory evoked potential (SSEPs) and clinical outcome upon intensive care unit discharge using the cerebral performance category (CPC) scale. RESULTS: We detected significant early HFO bursts in 146 patients and late HFO bursts in 95 patients. Only one of 27 unresponsive wakefulness patients had a late HFO burst amplitude above 70nV and all seventeen patients who died despite higher amplitudes died from non-neurological causes. CONCLUSIONS: High-frequency SSEP components can reliably be studied in comatose CA patients using standard equipment. SIGNIFICANCE: Late HFO burst amplitudes above 70nV largely exclude severe HE incompatible with regaining consciousness.

Non-invasive single-trial detection of variable population spike responses in human somatosensory evoked potentials.

OBJECTIVE: Somatosensory evoked potentials (SEPs) around 600 Hz ('σ-bursts') are correlates of cortical population spikes. Recently, single-trial σ-bursts were detected in human scalp EEG using 29-channel low-noise recordings in an electromagnetically shielded room. To achieve clinical applicability, this study aimed to establish a protocol using only 8 EEG channels in an unshielded environment and to quantify the variability of σ-bursts. METHODS: Median nerve SEPs were recorded in 10 healthy subjects using a custom-built low-noise EEG amplifier. A detection algorithm for single-trial σ-bursts was trained as combination of spatio-temporal filters and a non-linear classifier. The single-trial responses were probed for the presence of significant increases of amplitude and variability. RESULTS: Single-trial σ-burst detection succeeded with Detection Rates and Positive Predictive Values above 80% in subjects with high SNR. A significant inter-trial variability in the amplitudes of early low-frequency SEPs and σ-bursts could be demonstrated. CONCLUSIONS: Single-trial σ-bursts can be detected on scalp-EEG using only 8 EEG channels in an electromagnetically disturbed environment. The combination of dedicated hardware and detection algorithms allows quantifying and describing their variability. SIGNIFICANCE: The variability of population spikes in the human somatosensory cortex can be traced non-invasively in a clinical setting.

Recording human cortical population spikes non-invasively--An EEG tutorial.

BACKGROUND: Non-invasively recorded somatosensory high-frequency oscillations (sHFOs) evoked by electric nerve stimulation are markers of human cortical population spikes. Previously, their analysis was based on massive averaging of EEG responses. Advanced neurotechnology and optimized off-line analysis can enhance the signal-to-noise ratio of sHFOs, eventually enabling single-trial analysis. METHODS: The rationale for developing dedicated low-noise EEG technology for sHFOs is unfolded. Detailed recording procedures and tailored analysis principles are explained step-by-step. Source codes in Matlab and Python are provided as supplementary material online. RESULTS: Combining synergistic hardware and analysis improvements, evoked sHFOs at around 600 Hz ('σ-bursts') can be studied in single-trials. Additionally, optimized spatial filters increase the signal-to-noise ratio of components at about 1 kHz ('κ-bursts') enabling their detection in non-invasive surface EEG. CONCLUSIONS: sHFOs offer a unique possibility to record evoked human cortical population spikes non-invasively. The experimental approaches and algorithms presented here enable also non-specialized EEG laboratories to combine measurements of conventional low-frequency EEG with the analysis of concomitant cortical population spike responses.

Non-invasive single-trial EEG detection of evoked human neocortical population spikes.

QUESTION: Human high-frequency (>400 Hz) components of somatosensory evoked potentials (hf-SEPs), which can be recorded non-invasively at the scalp, are generated by cortical population spikes, as inferred from microelectrode recordings in non-human primates. It is a critical limitation to broader neurophysiological study of hf-SEPs in that hundreds of responses have to be averaged to detect hf-SEPs reliably. Here, we establish a framework for detecting human hf-SEPs non-invasively in single trials. METHODS: Spatio-temporal features were extracted from band-pass filtered (400-900 Hz) hf-SEPs by bilinear Common Spatio-Temporal Patterns (bCSTP) and then classified by a weighted Extreme Learning Machine (w-ELM). The effect of varying signal-to-noise ratio (SNR), number of trials, and degree of w-ELM re-weighting was characterized using surrogate data. For practical demonstration of the algorithm, median nerve hf-SEPs were recorded inside a shielded room in four subjects, spanning the hf-SEP signal-to-noise ratio characteristic for a larger population, utilizing a custom-built 29-channel low-noise EEG amplifier. RESULTS: Using surrogate data, the SNR proved to be pivotal to detect hf-SEPs in single trials efficiently, with the trade-off between sensitivity and specificity of the algorithm being obtained by the w-ELM re-weighting parameter. In practice, human hf-SEPs were detected non-invasively in single trials with a sensitivity of up to 99% and a specificity of up to 97% in two subjects, even without any recourse to knowledge of stimulus timing. Matching with the results of the surrogate data analysis, these rates dropped to 62-79% sensitivity and 18-31% specificity in two subjects with lower SNR. CONCLUSIONS: Otherwise buried in background noise, human high-frequency EEG components can be extracted from low-noise recordings. Specifically, refined supervised filter optimization and classification enables the reliable detection of single-trial hf-SEPs, representing non-invasive correlates of cortical population spikes. SIGNIFICANCE: While low-frequency EEG reflects summed postsynaptic potentials, and thereby neuronal input, we suggest that high-frequency EEG (>400 Hz) can provide non-invasive access to the unaveraged output of neuronal computation, i.e., single-trial population spike activity evoked in the responsive neuronal ensemble.

Are high-frequency (600 Hz) oscillations in human somatosensory evoked potentials due to phase-resetting phenomena?

OBJECTIVE: Median nerve somatosensory evoked potentials (SEP) contain a brief oscillatory wavelet burst at about 600 Hz (σ-burst) superimposed on the initial cortical component (N20). While invasive single-cell recordings suggested that this burst is generated by increased neuronal spiking activity in area 3b, recent non-invasive scalp recordings could not reveal concomitant single-trial added-activity, suggesting that the SEP burst might instead be generated by phase-reset of ongoing high-frequency EEG. Here, a statistical model and exemplary data are presented reconciling these seemingly contradictory results. METHODS: A statistical model defined the conditions required to detect added-activity in a set of single-trial SEP. Its predictions were tested by analyzing human single-trial scalp SEP recorded with custom-made low-noise amplifiers. RESULTS: The noise level in previous studies did not allow to detect single-trial added-activity in the period concomitant with the trial-averaged σ-burst. In contrast, optimized low-noise recordings do reveal added-activity in a set of single-trials. CONCLUSIONS: The experimental noise level is the decisive factor determining the detectability of added-activity in single-trials. A low-noise experiment provided direct evidence that the SEP σ-burst is at least partly generated by added-activity matching earlier invasive single-cell recordings. SIGNIFICANCE: Quantitative criteria are provided for the feasibility of single-trial detectability of band-limited added-activity.

Event-related desynchronization of sensorimotor EEG rhythms in hemiparetic patients with acute stroke.

Previous neuroimaging studies based on neurovascular coupling have shown that stroke affects both, strength and spatial extent of brain activation during upper limb movements. Here, we investigated the sub-second amplitude dynamics of a direct neuronal measure, i.e., event-related desynchronization (ERD) of EEG oscillations during finger movements, in patients with acute cortical and subcortical stroke. Acute cortical strokes were found to decrease the ERD of alpha oscillations for the affected pericentral sensorimotor areas compared to a control group. Within the cortical stroke group, the affected hemisphere showed a smaller alpha-ERD compared to the unaffected hemisphere when each was contralateral to the acting hand. Furthermore, when cortical stroke patients moved their paretic hand, the ipsilateral (i.e., contralesional) alpha-ERD was stronger than the contralateral (ipsilesional) ERD. Interestingly, the alpha-ERD amplitude in a hemisphere with a cortical stroke was relatively well preserved for non-paretic hand movements compared to alpha-ERD amplitude for paretic hand movements. This finding provides a new perspective for assessing the rehabilitative potential, which could be utilized through training of the still responsive cortical network, e.g., via enforced use of the paretic hand.