Objective response detection in an electroencephalogram during somatosensory stimulation

Multi-channel and multi-harmonic analysis of Auditory Steady-State Response detection

The Auditory Steady-State Response (ASSR) is a type of auditory evoked potential (AEP) generated in the auditory system that can be automatically detected by means of objective response detectors (ORDs). ASSRs are usually registered on the scalp using electroencephalography (EEG). ORD are univariate techniques, i.e. only uses one data channel. However, techniques involving more than one channel - multi-channel objective response detectors (MORDs) - have been showing higher detection rate (DR) when compared to ORD techniques. When ASSR is evoked by amplitude stimuli, the responses could be detected by analyzing the modulation frequencies and their harmonics. Despite this, ORD techniques are traditionally applied only in its first harmonic. This approach is known as one-sample test. The q-sample tests, however, considers harmonics beyond the first. Thus, this work proposes and evaluates the use of q-sample tests using a combination of multiple EEG channels and multiple harmonics of the stimulation frequencies and compare them with traditional one-sample tests. The database used consists of EEG channels from 24 volunteers with normal auditory threshold collected following a binaural stimulation protocol by amplitude modulated (AM) tone with modulating frequencies near 80 Hz. The best q-sample MORD result showed an increase in DR of 45.25% when compared with the best one-sample ORD test. Thus, it is recommended to use multiple channels and multiple harmonics, whenever available.

Characterization of interstimulus interaction in the multiple auditory steady-state responses at high sound levels

Multiple auditory steady-state response (MASSR) is recommended to estimate hearing thresholds in difficult-to-test individuals. The multiple stimuli that evoke MASSR may present an interstimulus interaction (ISI) that is able to distort the generation of responses. No consensus exists on the effects of the ISI in MASSR when dealing with high sound level stimuli or cases of sensorineural hearing loss. This study investigated the effects of ISI on the amplitude and detectability of auditory steady-state responses, with a focus at and above 65 dB sound pressure level (SPL). Normal hearing (NH) and sensorineural hearing impaired (SNHI) adults were tested with different stimulus types [amplitude modulation (AM) One octave chirp (OC), and a weighted OC (WOC)], stimulus levels, and modalities (single or multiple stimuli). ISI typically attenuated response amplitude of a control stimulus caused by an interference stimulus one octave above the control stimulus. At and above 80 dB SPL, attenuations of around 50% decreased the number of detectable responses near SNHI thresholds, especially for OC and WOC. AM stimuli obtained a higher detection rate than OC and WOC when presented 10 dB above the behavioral hearing threshold of SNHI participants. Using OC in MASSR when assessing elevated thresholds might diminish accuracy on threshold estimation, and extend test duration.

Faster automatic ASSR detection using sequential tests

Objective: Objective Response Detection (ORD) can be used for auditory steady-state response (ASSR) detection. In conventional ORD methods, the statistical tests are applied at the end of data collection ('single-shot tests'). In sequential ORD methods, statistical tests are applied repeatedly, while data is being collected. However, repeated testing can increase False Positive (FP) rates. One solution is to infer that response is present only after the test remains significant for a predefined number of consecutive detections (NCD). Thus, this paper describes a new method for finding the required NCD that control the FP rate for ASSR detection.Design: NCD values are estimated using Monte Carlo simulations.Study sample: ASSR signals were recorded from 8 normal-hearing subjects.Results: The exam time was reduced by up to 38.9% compared to the single-shot test with loss of approximately 5% in detection rate. Alternatively, lower gains in time were achieved for a smaller (non-significant) loss in detection rate. The FP rates at the end of the test were kept at the nominal level expected (1%).Conclusion: The sequential test strategy with NCD as the stopping criterion can improve the speed of ASSR detection and prevent higher than expected FP rates.

A bayesian approach to the spectral F-Test: Application to auditory steady-state responses

BACKGROUND AND OBJECTIVE

Auditory steady-state responses (ASSRs) represent an objective method used in clinical practice to assess hearing thresholds. The steady-state nature of these signals allows response detection by means of statistical techniques in the frequency domain as spectral F-test. This objective response detection (ORD) compares the power of the response bin against the power of the neighboring frequency noise bins. Most ORD algorithms are based on the Neyman-Pearson approach to the hypothesis test provided that the likelihood ratio test is the most powerful test for a given significance level alpha (also called Type I error). On the other hand, the Bayesian approach allows the inclusion of prior information in the model and enables the updating of this information with posterior knowledge. This approach, however, has not been explored with respect to ORD techniques, thus enabling the exploration of new paradigms, which may contribute to this field of study, especially in terms of the time required for response detection. The aim of this study is to use the Bayesian approach in the implementation of the spectral F-test for application to ASSRs.

METHODS

Monte Carlo simulations were performed to evaluate Neyman-Pearson and Bayesian detectors' performances with the spectral F-test as a function of the signal-to-noise ratio. Then, the two detectors were applied to ASSR recordings of nine normal-hearing individuals subjected to amplitude-modulated tones of various intensities.

RESULTS

Both simulation and ASSR data analyses showed that among the scenarios analyzed, the most promising case was that in which the lowest possible values for the a priori probability were selected for the null hypothesis (H0), allowing detection at low signal-to-noise ratios. The worst performance occurred when the a priori probabilities for both hypotheses were equal. The ASSR data also showed that higher stimulus intensity led to better performance and faster detection due to improvements in the signal-to-noise ratio.

CONCLUSIONS

The a priori probabilities can affect the Bayesian detector's performance, directly impacting the time needed to identify responses. The parallel behaviors observed between the performances of both approaches showed that the Bayesian detector can achieve its ideal performance at lower signal-to-noise ratios compared to the optimal performance of the Neyman-Pearson detector, reflecting the promising applicability of the Bayesian approach to evoked potentials.

The Convolutional Group Sequential Test: Reducing Test Time for Evoked Potentials

When using a statistical test for automatically detecting evoked potentials, the number of stimuli presented to the subject (the sample size for the statistical test) should be specified at the outset. For evoked response detection, this may be inefficient, i.e., because the signal-to-noise ratio (SNR) of the response is not known in advance, the user would usually err on the cautious side and use a relatively high number of stimuli to ensure adequate statistical power. A more efficient approach is to apply the statistical test repeatedly to the accumulating data over time, as this allows the test to be stopped early for the high SNR responses (thus reducing test time), or later for the low SNR responses. The caveat is that the critical decision boundaries for rejecting the null hypothesis need to be adjusted if the intended type-I error rate is to be obtained. This study presents an intuitive and flexible method for controlling the type-I error rate for sequentially applied statistical tests. The method is built around the discrete convolution of truncated probability density functions, which allows the null distribution for the test statistic to be constructed at each stage of the sequential analysis. Because the null distribution remains tractable, the procedure for finding the stage-wise critical decision boundaries is greatly simplified. The method also permits data-driven adaptations (using data from previous stages) to both the sample size and the statistical test, which offers new opportunities to speed up testing for evoked response detection.

Detecting neuromagnetic synchrony in the presence of noise

BACKGROUND

Synchrony between neuroelectric oscillations in distant brain areas is currently used as an indicator of functional connectivity between the involved neural substrates. Coherence measures, which quantify synchrony, are affected by concurrent brain activities, commonly subsumed as noise.

NEW METHOD

Using Monte-Carlo simulation, we analysed the properties of circular statistics and how those are affected by noise. We considered three different models of neuroelectric signal generation, which are an additive model, phase-reset, and reciprocal phase-interaction. Using the receiver-operating characteristic method, we compared the performances of currently implemented algorithms for coherence detection such as phase-coherence or phase-locking factor, magnitude-squared coherence, and phase-lagging index, all based on circular statistics, and a more general approach to synchrony, using measures of mutual information. We compared inter-trial coherence as a method for signal detection with coherence between multiple sources as measure of source interaction and connectivity.

RESULTS

Charts of performance characteristics showed that the choice of methods depend on the underlying signal generation model. Detection of coherence requires in general a higher signal-to-noise ratio than detection of the signal itself, and again, the difference in performance depends strongly on the underlying model of signal generation.

COMPARISON WITH EXISTING METHODS

Previous comparisons of the performances of different algorithms for signal detection and coherence have not considered systematically the underlying neural generation mechanisms.

CONCLUSION

Detection of coherence generated by additive signals or a phase-reset requires largely higher signal-to-noise ratio compared to signal detection. Only in case of true phase interaction, signal detection and coherence measures are similarly sensitive.

Validation in principal components analysis applied to EEG data

The well-known multivariate technique Principal Components Analysis (PCA) is usually applied to a sample, and so component scores are subjected to sampling variability. However, few studies address their stability, an important topic when the sample size is small. This work presents three validation procedures applied to PCA, based on confidence regions generated by a variant of a nonparametric bootstrap called the partial bootstrap: (i) the assessment of PC scores variability by the spread and overlapping of “confidence regions” plotted around these scores; (ii) the use of the confidence regions centroids as a validation set; and (iii) the definition of the number of nontrivial axes to be retained for analysis. The methods were applied to EEG data collected during a postural control protocol with twenty-four volunteers. Two axes were retained for analysis, with 91.6% of explained variance. Results showed that the area of the confidence regions provided useful insights on the variability of scores and suggested that some subjects were not distinguishable from others, which was not evident from the principal planes. In addition, potential outliers, initially suggested by an analysis of the first principal plane, could not be confirmed by the confidence regions.

Some thoughts on the interpretation of steady-state evoked potentials

Steady-state evoked potentials are popular due to their easy analysis in frequency space and the availability of methods for objective response detection. However, the interpretation of steady-state responses can be challenging due to their origin as a sequence of responses to single stimuli. In the present paper, issues of signal extinction and some characteristics of higher harmonics are illustrated based on simple model data for those readers who do not regularly hobnob with frequency-space representations of data. It is important to realize that the absence of a steady-state response does not prove the lack of neural activity. For the same underlying reasons, namely constructive and destructive superposition of individual responses, comparisons of amplitudes between experimental conditions are prone to inaccuracies. Thus, before inferring physiology from steady-state responses, one should consider an alternative explanation in terms of signal composition.

Topographic distribution of the tibial somatosensory evoked potential using coherence

The objective of the present study was to determine the adequate cortical regions based on the signal-to-noise ratio (SNR) for somatosensory evoked potential (SEP) recording. This investigation was carried out using magnitude-squared coherence (MSC), a frequency domain objective response detection technique. Electroencephalographic signals were collected (International 10-20 System) from 38 volunteers, without history of neurological pathology, during somatosensory stimulation. Stimuli were applied to the right posterior tibial nerve at the rate of 5 Hz and intensity slightly above the motor threshold. Response detection was based on rejecting the null hypothesis of response absence (significance level alpha= 0.05 and M = 500 epochs). The best detection rates (maximum percentage of volunteers for whom the response was detected for the frequencies between 4.8 and 72 Hz) were obtained for the parietal and central leads mid-sagittal and ipsilateral to the stimulated leg: C4 (87%), P4 (82%), Cz (89%), and Pz (89%). The P37-N45 time-components of the SEP can also be observed in these leads. The other leads, including the central and parietal contralateral and the frontal and fronto-polar leads, presented low detection capacity. If only contralateral leads were considered, the centro-parietal region (C3 and P3) was among the best regions for response detection, presenting a correspondent well-defined N37; however, this was not observed in some volunteers. The results of the present study showed that the central and parietal regions, especially sagittal and ipsilateral to the stimuli, presented the best SNR in the gamma range. Furthermore, these findings suggest that the MSC can be a useful tool for monitoring purposes.

Use of magnitude-squared coherence to identify the maximum driving response band of the somatosensory evoked potential

The present study proposes to apply magnitude-squared coherence (MSC) to the somatosensory evoked potential for identifying the maximum driving response band. EEG signals, leads [Fpz'-Cz'] and [C3'-C4'], were collected from two groups of normal volunteers, stimulated at the rate of 4.91 (G1: 26 volunteers) and 5.13 Hz (G2: 18 volunteers). About 1400 stimuli were applied to the right tibial nerve at the motor threshold level. After applying the anti-aliasing filter, the signals were digitized and then further low-pass filtered (200 Hz, 6th order Butterworth and zero-phase). Based on the rejection of the null hypothesis of response absence (MSC(f) > 0.0060 with 500 epochs and the level of significance set at a = 0.05), the beta and gamma bands, 15-66 Hz, were identified as the maximum driving response band. Taking both leads together ("logical-OR detector", with a false-alarm rate of a = 0.05, and hence a = 0.0253 for each derivation), the detection exceeded 70% for all multiples of the stimulation frequency within this range. Similar performance was achieved for MSC of both leads but at 15, 25, 35, and 40 Hz. Moreover, the response was detected in [C3'-C4'] at 35.9 Hz and in [Fpz'-Cz'] at 46.2 Hz for all members of G2. Using the "logical-OR detector" procedure, the response was detected at the 7th multiple of the stimulation frequency for the series as a whole (considering both groups). Based on these findings, the MSC technique may be used for monitoring purposes.

Low-frequency oscillations in human tibial somatosensory evoked potentials

Oscillatory cerebral electric activity has been related to sensorial and perceptual-cognitive functions. The aim of this work is to investigate low frequency oscillations (<300 Hz), particularly within the gamma band (30-110 Hz), during tibial stimulation. Twenty-one volunteers were subjected to 5 Hz stimulation by current pulses of 0.2 ms duration and the minimum intensity to provoke involuntary twitch. EEG signals without (spontaneously) and during stimulation were recorded at primary somatosensory area. A time-frequency analysis indicated the effect of the stimulus artifact in the somatosensory evoked potential (SEP) frequencies up to 5 ms after the stimulus. The oscillations up to 100 Hz presented the highest relative power contribution (approximately 99%) for the SEP and showed difference (p<0.01) from the frequencies of the spontaneously EEG average. Moreover, the range 30-58 Hz was identified as the band with the highest contribution for the tibial SEP morphology (p<0.0001).

Objective detection of evoked potentials using a bootstrap technique

Evoked potentials are usually evaluated subjectively, by visual inspection, and considerable differences between interpretations can occur. Objective, automated methods are normally based on calculating one (or more) parameters from the data, but only some of these techniques can provide statistical significance (p-values) for the presence of a response. In this work, we propose a bootstrap technique to provide such p-values, which can be applied to a wide variety of parameters. The bootstrap method is based on randomly resampling (with replacement) the original data and gives an estimate of the probability that the response obtained is due to random variation in the data rather than a physiological response. The method is illustrated using auditory brainstem responses (ABRs) to detecting hearing thresholds. The flexibility of the approach is illustrated, showing how it can be used with different parameters, numbers of stimuli and with user-defined false-positive rates. The bootstrap method provides a new, simple and yet powerful means of detecting evoked potentials, which is very flexible and readily adapted to a wide variety of signal parameters.

Avoiding spectral leakage in objective detection of auditory steady-state evoked responses in the inferior colliculus of rat using coherence

Local field potentials (LFP) are bioelectric signals recorded from the brain that reflect neural activity in a high temporal resolution. Separating background activity from that evoked by specific somato-sensory input is a matter of great clinical relevance in neurology. The coherence function is a spectral coefficient that can be used as a detector of periodic responses in noisy environments. Auditory steady-state responses to amplitude-modulated tones generate periodic responses in neural networks that may be accessed by means of coherence between the stimulation signal and the LFP recorded from the auditory pathway. Such signal processing methodology was applied in this work to evaluate in vivo, anaesthetized Wistar rats, activation of neural networks due to single carrier sound stimulation frequencies, as well as to evaluate the effect of different modulating tones in the evoked responses. Our results show that an inappropriate choice of sound stimuli modulating frequencies can compromise coherence analysis, e.g. misleading conclusions due to mathematical artefact of signal processing. Two modulating frequency correction protocols were used: nearest integer and nearest prime number. The nearest prime number correction was successful in avoiding spectral leakage in the coherence analysis of steady-state auditory response, as predicted by Monte Carlo simulations.

Evaluating the entrainment of the alpha rhythm during stroboscopic flash stimulation by means of coherence analysis

Two major conflicting hypotheses propose that alpha rhythm activity should be either the output of a linear filter having a white noise as input or reflect the output of a nonlinear oscillator. External stimulation can be employed to test for nonlinearity in alpha genesis, since an entrainment of such rhythmic activity (shift in the alpha peak) could only be explained by nonlinear relationships. Flash photic stimulation has been used to investigate such entrainment. Nevertheless, only entrainments due to the second harmonic of the stimulation could be suitably measured. Aiming at overcoming this limitation, a coherence-based technique is proposed for evaluating the strength of responses due to rhythmic stimulation. It was applied to the occipital EEG derivations of 12 normal subjects during stroboscopic stimulation. Entrainment of alpha rhythm by the second harmonic of the stimulation occurred in 75% of the subjects, whilst no spectral shifts were observed for the remained that exhibited broadband alpha peak at rest. However, stimulating with fundamental frequency close to that peak led to entrainment in all subjects. These differences in the degree of synchronization due to stimulation at the first and second harmonics should reflect complex nonlinear mechanisms in alpha genesis.

A Matrix-Based Algorithm for Estimating Multiple Coherence of a Periodic Signal and Its Application to the Multichannel EEG During Sensory Stimulation

The coherence between the stimulation signal and the electroencephalogram (EEG) has been used in the detection of evoked responses. The detector's performance, however, depends on both the signal-to-noise ratio (SNR) of the responses and the number of data segments (M) used in coherence estimation. In practical situations, when a given SNR occurs, detection can only be improved by increasing M and hence the total data length. This is particularly relevant when monitoring is the objective. In the present study, we propose a matrix-based algorithm for estimating the multiple coherence of the stimulation signal taking into account a set of N EEG channels as a way of increasing the detection rate for a fixed value of M. Monte Carlo simulations suggest that thresholds for such multivariate detector are the same as those for multiple coherence of Gaussian signals and that using more than six signals is not advisable for improving the detection rate with M = 10. The results with EEG from 12 normal subjects during photic stimulation at 10 Hz showed a maximum detection for N greater than 2 in 58% of the subjects with M = 10, and hence suggest that the proposed multivariate detector is valuable in evoked responses applications.

Multi-channel evoked response detection using only phase information

The phase consistency of contiguous segments of the electroencephalogram (EEG) has been used in the detection of evoked responses to rhythmic stimulation. One of such techniques is the component synchrony measure (CSM), which is often used since the threshold for the detection task is easily obtained based on the estimates of asymptotic sample distribution. In this work we investigated the appropriateness of such thresholds for practical number of segments (M). The performance of CSM was next evaluated by Monte Carlo simulations with different signal-to-noise ratios (SNR) and values of M, and the results, compared with those for the magnitude-squared coherence. A way of improving the detection with CSM was also proposed, by suggesting the estimation taking into account the mean phase angle of a set of N signals. This multivariate detector was evaluated in simulations and an illustration of the technique was also given with the EEG of 14 subjects during photic stimulation. In simulated signals with equal SNR, the detection rate with this multivariate measure increased with N. The application to EEG data lead to similar results in 70% of the subjects, which suggests that improvements might be expected when more signals are available to detect evoked responses in EEG.

Improving the detection of evoked responses to periodic stimulation by using multiple coherence--application to EEG during photic stimulation

The coherence between the stimulation signal and the electroencephalogram (EEG) has been used in the detection of evoked responses. However the detector's performance depends on both the signal-to-noise ratio (SNR) of the responses and the number of data segments (M) used in coherence estimation. In this work, a technique for detecting evoked responses was developed based on the extension to the multivariate case of this coherence. Thus, instead of using the EEG collected at a unique region, the estimation is proposed using two EEG derivations. As for the univariate case, this multiple coherence is independent of the stimulation signal. In addition, considering equal SNR in both signals, the detection rate with this multiple coherence is always greater than that one using only one signal. This was verified in Monte Carlo simulations, which also showed that a superior performance is still expected in practical situations, when a smaller SNR is found in the second signal. The results with EEG from 12 normal subjects during photic stimulation confirm this better performance. Since the proposed technique allows a higher detection rate without the need of increasing M, it permits evoked responses to be detected faster, which is very useful during monitored surgeries.

Coherence between one random and one periodic signal for measuring the strength of responses in the electro-encephalogram during sensory stimulation

Coherence between a pulse train representing periodic stimuli and the EEG has been used in the objective detection of steady-state evoked potentials. This work aimed to quantify the strength of the stimulus responses based on the statistics of coherence estimate between one random and one periodic signal, focusing on the confidence limits and power of significance tests in detecting responses. To detect the responses in 95% of cases, a signal-to-noise ratio of about -7.9 dB was required when using 48 windows (M) in the coherence estimation. The ratio, however, increased to -1.2 dB when M was 12. The results were tested in Monte Carlo simulations and applied to EEGs obtained from 14 subjects during visual stimulation. The method showed differences in the strength of responses at the stimulus frequency and its harmonics, as well as variations between individuals and over cortical regions. In contrast to those from the parietal and temporal regions, results for the occipital region gave confidence limits (with M = 12) that were above zero for all subjects, indicating statistically significant responses. The proposed technique extends the usefulness of coherence as a measure of stimulus responses and allows statistical analysis that could also be applied usefully in a range of other biological signals.

A statistical technique for measuring synchronism between cortical regions in the EEG during rhythmic stimulation

The coherence function has been widely applied in quantifying the degree of synchronism between electroencephalogram (EEG) signals obtained from different brain regions. However, when applied to investigating synchronization resulting from rhythmic stimulation, misleading results can arise from the high correlation of background EEG activity. We, thus propose a modified measure, which emphasizes the synchronized stimulus responses and reduces the influence of the spontaneous EEG activity. Critical values for this estimator are derived and tested in Monte Carlo simulations. The effectiveness of the method is illustrated on data recorded from 12 young normal subjects during rhythmic photic stimulation.