Leonardo Bonato Felix

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 detectors 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.

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.

Coherence estimate between a random and a periodic signal: Bias, variance, analytical critical values, and normalizing transforms

The present work deals with recent results on the sampling distribution of the magnitude-squared coherence (also called just coherence) estimate between a random (Gaussian) and a periodic signal, in order to obtain analytical critical values, alternative expressions for the probability density function (PDF) as well as the variance and bias of the estimate. A comparison with the more general case of coherence estimation when both signals are Gaussian was also provided. The results indicate that the smaller the true coherence (TC) values the closer both distributions become. The behaviour of variance and bias as a function of the number of data segments and the TC is similar for both coherence estimates. Additionally, the effect of a normalizing function (Fishers z transform) in the coherence estimated between a random and a periodic signal was also evaluated and normality has been nearly achieved. However, the variance was less equalized in comparison with coherence estimate between two Gaussian signals.

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.

Comments on “Sums, Products, and Ratios of Non-Central Beta Variables” by Saralees Nadarajah

Brief comments on “Sums, Products, and Ratios of Non-Central Beta Variables” by Saralees Nadarajah, which appeared in Communications in Statistics – Theory and Methods, Volume 34, Issue 1, pp. 89–100.

Post-processing of auditory steady-state responses to correct spectral leakage.

Auditory steady-state responses (ASSRs) are electrical manifestations of brain due to high rate sound stimulation. These evoked responses can be used to assess the hearing capabilities of a subject in an objective, automatic fashion. Usually, the detection protocol is accomplished by frequency-domain techniques, such as magnitude-squared coherence, whose estimation is based on the fast Fourier transform (FFT) of several data segments. In practice, the FFT-based spectrum may spread out the energy of a given frequency to its side bins and this escape of energy in the spectrum is called spectral leakage. The distortion of the spectrum due to leakage may severely compromise statistical significance of objective detection. This work presents an offline, a posteriori method for spectral leakage minimization in the frequency-domain analysis of ASSRs using coherent sampling criterion and interpolation in time. The technique was applied to the local field potentials of 10 Wistar rats and the results, together with those from simulated data, indicate that a leakage-free analysis of ASSRs is possible for any dataset if the methods showed in this paper were followed.

Improving the detection of evoked responses to periodic stimulation by using bivariate local spectral F-test – Application to EEG during photic stimulation

The spectral local F-test has been applied for detecting evoked responses to rhythmic stimulation that are embedded in the ongoing electroencephalogram (EEG). Based on the sampling distribution of a flat spectrum at the neighbourhood of the stimulation frequency, spectral peaks in an EEG signal that are due to the stimulation may be readily assessed. Nevertheless, the performance of the technique is strongly affected by both the signal-to-noise ratio (SNR) of the responses and the number of data segments used in the estimation. The present work aims at both deriving and evaluating a multivariate extension of local F-test by including the EEG collected at a second distinct derivation. The detection rate with this multivariate detector was found to be greater than that using a single channel in case of equal SNR in both signals. Monte Carlo simulation results showed that the probability of detection with this new detector saturates for signal-to-noise ratios above 12 dB and indicated a greater detection rate in practical situations, even when smaller SNR-values are found in the added signal (e.g. 5 dB for 16 neighbouring frequencies used in the estimation). The technique was next applied to the EEG from 12 subjects during intermittent, photic stimulation leading to superior performance in comparison with the univariate local F-test. Since a higher detection rate with the proposed technique is achieved without the need of increasing the number of data segments, it allows evoked responses to be detected faster, once the same detection rate may be accomplished with less segments. This might be useful in clinical practice.

Classification of auditory selective attention using spatial coherence and modular attention index

BACKGROUND AND OBJECTIVE Brain-Computer Interfaces (BCIs) based on auditory selective attention have been receiving much attention because i) they are useful for completely paralyzed users since they do not require muscular effort or gaze and ii) focusing attention is a natural human ability. Several techniques - such as recently developed Spatial Coherence (SC) - have been proposed in order to optimize the BCI procedure. Thus, this work aims at investigating and comparing two strategies based on spatial coherence detection: contralateral and modular classifiers. The latter is a new method using modular attention index. The new classifier was developed to implement an auditory BCI where a volunteer makes binary choices using selective attention under the amplitude-modulated tones stimulation. METHODS Contralateral and modular classifiers were applied to the electroencephalogram (EEG) recorded from 144 subjects under the BCI protocol. The best set of parameters (carriers of the stimulus, channels and trials of signal) for this BCI was investigated taking into consideration the hit rate and the information transfer rate. RESULTS The best result obtained using the modular classifier was a hit rate of 91.67% and information transfer rate of 6.74 bits/min using 0.5 kHz/4.0 kHz as stimuli and three windows (5.10 sec of EEG signal). These results were obtained with five electrodes (C3, P3, F8, P4, O2) using exhaustive search to identify regions with greater coherence. CONCLUSION The modular classifier - using electroencephalogram channels from the central, frontal, occipital and parietal areas - improves the performance of auditory BCIs based on selective attention.

Evaluation of the movement imagination training using the principal component analysis and magnitude-squared coherence as extractors of features

This work investigates the Movement Imagination training using Principal Component Analysis (PCA) and Magnitude-Squared Coherence (MSC) as features extractor. The characteristics were extracted by using the Delta band (0.1-2 Hz), Alpha band (8-13 Hz) and Beta band (14-30 Hz) and the classifier was Multilayer Perceptron (MLP). Thus., the electroencephalogram (EEG) from five healthy subjects was recorded in the derivations C1, C2, C3, C4, C5, C6 and Cz (10-10 International System). The average hit rate in classification were 63.92 0/0, 71.31 0/0, 73.86 0/0, 83.31 0/0, 81.09 % and 93.43 % to 1st, 2nd, 3rd, 4th, 5th and 6th stages of training, respectively. Therefore., the results show the training increased the classifier hit rate using PCA and MSC as feature extractor.

An online coherence-based BCI for driving a mechanical arm

The present work describes the simulation to be used in the control of a mechanical arm using a Visual Evoked Potential to drive a Brain Computer Interface system. The signal processing and classifying was done using the Multiple Coherence K2N. The proposed classifier was able to detect the difference between four different frequencies presented at the same time. The Multiple Coherence classifier had a hit rate of 95%.