Selective averaging of the intracerebral click evoked responses in man: an improved method of measuring latencies and amplitudes

Single-trial event-related potentials with wavelet denoising

The application of a recently proposed denoising implementation for obtaining event-related potentials (ERPs) at the single-trial level is shown. We study its performance in simulated data as well as in visual and auditory ERPs. For the simulated data, the method gives a significantly better reconstruction of the single-trial event-related responses in comparison with the original data and also in comparison with a reconstruction based on conventional Wiener filtering. Moreover, with wavelet denoising we obtain a significantly better estimation of the amplitudes and latencies of the simulated ERPs. For the real data, the method clearly improves the visualization of both visual and auditory single-trial ERPs. This allows the calculation of better averages as well as the study of systematic or unsystematic variations between trials. Since the method is fast and parameter free, it could complement the conventional analysis of ERPs.

Enhanced brainstem and cortical evoked response amplitudes: single-trial covariance analysis

The purpose of the present study was to develop analytic procedures that improve the definition of sensory evoked response components. Such procedures could benefit all recordings but would especially benefit difficult recordings where many trials are contaminated by muscle and movement artifacts. First, cross-correlation and latency adjustment analyses were applied to the human brainstem frequency-following response and cortical auditory evoked response recorded on the same trials. Lagged cross-correlation functions were computed, for each of 17 subjects, between single-trial data and templates consisting of the sinusoid stimulus waveform for the brainstem response and the subject's own smoothed averaged evoked response P2 component for the cortical response. Trials were considered in the analysis only if the maximum correlation-squared (r2) exceeded .5 (negatively correlated trials were thus included). Identical correlation coefficients may be based on signals with quite different amplitudes, but it is possible to assess amplitude by the nonnormalized covariance function. Next, an algorithm is applied in which each trial with negative covariance is matched to a trial with similar, but positive, covariance and these matched-trial pairs are deleted. When an evoked response signal is present in the data, the majority of trials positively correlate with the template. Thus, a residual of positively correlated trials remains after matched covariance trials are deleted. When these residual trials are averaged, the resulting brainstem and cortical responses show greatly enhanced amplitudes. This result supports the utility of this analysis technique in clarifying and assessing evoked response signals.

Cross-correlation and latency compensation analysis of click-evoked and frequency-following brain-stem responses in man

Cross-correlation (CC) and latency compensation (LC) analysis were applied to the human click-evoked brain-stem auditory evoked response (BAER) and the brain-stem frequency-following response (FFR). FFRs were elicited by pure tone stimuli (230 Hz and 460 Hz) or by complex tones derived from the sum of 3rd (920 Hz), 4th (1150 Hz), and 5th (1380 Hz) harmonics of the missing 230 Hz fundamental. The lower and upper harmonics always began in sine phase, while the middle harmonic varied in starting phase, resulting in harmonically complex stimuli with differing amplitude and phase patterns. Cross-correlations were computed between individual trials and a wave form template (smoothed wave V for BAER, pure tone stimulus sinusoids for FFR). Trials were included in the analysis only if values of r2 exceeded 0.5 (negative values of r were thus included, which controlled for the chance occurrence of positive correlations). Although brain-stem recordings are noisy, requiring as many as 1000 stimuli/average, correlation analysis consistently identified more positive than negative trials (approximately 2:1 ratio). Trials were also deleted if the lag associated with the selected r2 was at the maximum shift position ('extreme lag'). Averaging trials that satisfied the correlation and lag criteria led to sizeable enhancement of BAER (mean = 114%) and FFR (mean = 68% for 230 Hz stimulus) amplitudes. LC analysis resulted in additional, albeit smaller, increases in amplitude (approximately 10%). FFRs to harmonically complex stimuli were characterized by a clear periodicity at the missing fundamental frequency (230 Hz). However, amplitudes varied according to the modulation depth of the stimulus and, in certain cases, actually exceeded that of the FFR response to a 230 Hz pure tone. The results demonstrate the effectiveness of cross-correlation and, to a lesser degree, latency compensation analysis, applied to two classes of brain-stem potentials. It is anticipated that such techniques will prove useful in the study of auditory signal processing at the level of the brain-stem.

Real-time reconstruction of evoked potentials using a new two-dimensional filter method

Standard techniques for evoked potential recording extract a stimulus-locked event from accompanying noise by averaging a large number of sequentially obtained responses. This approach is valid only to the extent that the nervous system's electrical response to successive stimuli is identical. The technique is suboptimal for recording unstable evoked potentials which vary with the subject's state and attention. Similarly, standard methods are suboptimal for efficiently analyzing rapid changes such as may be seen in the operating room. We developed an evoked potential recording method that reconstructs the individual evoked responses (or small subaverages of evoked potentials) to a series of stimuli. First, the raw data from an entire series of one to several hundred responses are recorded digitally. Using a frequency domain two-dimensional filter, the data are then filtered twice, once along the data sequence axis for each trial, and again along the cross-trial sequence axis for comparable frequency coefficients in sequential trials. The reconstructed filtered evoked potentials are plotted, with successive responses stacked for easy tracking of component changes.

Detection of reaction time by an adaptive filter based on the least squares fit

An adaptive filter based on the least squares fit between a template and individual records was developed. It has been applied to compute reaction times of a voluntary movement. Its performance was tested by comparing these reaction times with the latency of the EMG responses. Different properties of the least squares fit technique and Woody's filter technique (Woody 1967) are discussed.

Averaging, selective averaging and latency-corrected averaging

Although latency-corrected averaging and selective averaging represent frequency used alternatives to the time-locked averaging, little is known about the changes of the signal-to-noise ratio (SNR) associated with their application. For this reason, visual evoked responses (VERs) to blank and checkerboard patterned flashes of light (5.6 cd/m2) were recorded in 15 subjects using a referential (Oz - A1 + 2) and a bipolar (Oz - Pz) derivation, and interstimulus intervals (ISI) of 1.4s and 2.4s; SNRs were calculated for all combinations. With time-locked averaging (referential derivation) the mean SNR of single VERs was 0.35 to blank stimuli and 0.68 to patterned stimuli. With latency-corrected averaging, the means SNR rose to 0.56 (0.74), and with selective averaging it increased to 0.81 (1.00). A combined procedure of selective and latency-corrected averaging yielded the highest mean SNR values of single VERs: 0.98 (1.21). Increasing ISIs from 1.4 - 2.4s did not change the results; the mean SNR values were always higher in bipolar than in referential derivations. The selective, latency-corrected averaging produced higher SNRs, and the group mean number of detectable single VERs (62% of the single VERs to blank stimuli, and 74% of the single VERs to patterned stimuli) and the latency variability of the single VERs: -0.33 +/- 14.5 ms (blank stimulation) and -0.48 +/- 13.96 ms (patterned stimulation).

A posteriori time-varying filtering of averaged evoked potentials. I. Introduction and conceptual basis

This paper forms a preface and introduction to a new method for the estimation of evoked potentials: a posteriori time-varying filtering. A simple evoked potential model, consisting of a transient signal and additive noise, is discussed and the underlying assumptions explicitly formulated. Assuming this model, the problem of estimating the signal from an ensemble is considered from the statistical and communication engineering point of view, along with a brief survey of he pertinent literature. It is explained why ensemble averaging, in general, does not provide the best estimate in the mean-square error sense. After a summary of the controversial aspects of time-invariant "a posteriori "Wiener' filtering", it is indicated how that method can be generalized to a time-varying counterpart, which is able to handle the essentially transient character of evoked potential waveforms. Finally, the new method is presented on a conceptual level and its application illustrated by examples.

N1-P2 correlates of reaction time at the single-trial level

Using a template-matching pattern recognition technique evaluated in an earlier paper, the amplitude and latency of the N1-P2 wave of the vertex visual evoked potential was established on each trial of a reaction time task. This single-trial physiological information correlated meaningfully with trial-by-trial changes in performance, providing further indication of the validity and practicability of a single-trial approach to the analysis of human EP data.

Statistical detection of individual evoked responses: an evaluation of Woody's adaptive filter

The practical performance of the adaptive filter developed by Woody (1967) and Harris and Woody (1969) is assessed in terms of its ability to distinguish evoked responses from the background EEG. It is concluded that selection of good initial template, in this case the Averaged ER, renders iterations beyond the first doubtful validity in the majority of subjects. It is stressed that where iterations are pursued care should be taken over the choice of a suitable statistic to index the process. Finally, given the efficacy of the cavariance function in detecting the grosser features of ER morphology in individual responses, a single trial approach to the analysis of ERs is recommended.

Comparison of Wiener filtering and selective averaging of evoked potentials

The application of the Wiener filter to the estimation of evoked potentials is criticized, and this method is compared with an a posteriori selective averaging method. It is shown that Wiener filtering may cause information loss for certain types of evoked potentials, since the transient evoked response components of the brain are of damped oscillatory character and are not stationary signals. The selective averaging method is briefly described and suggested to obtain consistent, dependable and more descriptive averaged evoked potentials (AEPs) of a brain structure for well-defined waking and sleep stages. The stated arguments are supported by a comparative representation of AEPs obtained from the cat inferior colliculus by means of Wiener filtering and selective averaging.

Variability of cortical evoked responses in man related to slow wave activity

Acoustically evoked responses were recorded from subdural and intracerebral electrodes chronically implanted in man. It was found that during slow wave activity especially the negative component at about 100 ms latency varies in form and amplitude. If the stimuli were presented during increasing cortical negativity, the amplitude of the 100 ms component of the evoked responses was generally larger than in the case of stimuli presented during decreasing negativity.