A Method for Tracking the Time Evolution of Steady-State Evoked Potentials

Neural entrainment is the synchronization of the neural activity to the periodicity of the sensory stimuli. This synchronization generates the steady-state evoked potential, that is, oscillations in the electroencephalogram phase-locked to the sensory stimuli. The classic interpretation of the amplitude of the steady-state evoked responses, assumes a constant neural response phase-locked to the stimulus, plus a random background noise not related to the stimulus.

The stereotypical response can be obtained averaging over repeated presentation of the stimulus. This approach ignores the dynamics of the response, as in the case of the evoked potential adaptation, elicited by prolonged exposure to the stimulus. In animal models, the auditory steady-state response generated in the cortical brain regions, and adds to the continuous presentation of amplitude-modulated tones.

In humans, it has been recently demonstrated that the power of the fundamental frequency of steady-state visually invoked potential is stationary only in 30%of the subjects. When the focus of the research is the dynamics of entrainment, we can assume that the temporal evolution of the response will be the same in different independent experimental runs. Therefore, the average of the signal in every epoch across independent runs we provide an accurate representation of the long-term dynamics of the oscillatory response.

Based on that assumption, we have developed a method to characterize the time evolution of the steady-state response. The method consists of acquiring several recordings of the same experimental condition, following instead of averaging subsequent epochs within the recordings. Epochs which correspond to the same time window in the different recordings are averaged.

In this study we provide a detailed description of the method, using steady-state visually evoked potentials as an example of a response. However, the the methodology can be used to analyze the steady-state responses of other sensory stimuli. Finally, we present the advantages and drawbacks of the methodology, based on the comparison with single-trial methods use it to analyze the neural entrainment.

Welcome the subject. Invite the subject to talk in a friendly atmosphere, to explain him or her the aims and relevance of this study. Provide a description of relevant technical details.

Answer all his or her questions thoroughly. Explicitly mention that she or he is allowed to interrupt the experiment session at any time if desired. Ask the volunteer to read the Subject Informed Consent and sign the corresponding form.

Clean the scalp with ethanol, a solution at 95%to remove the layer of dead skin cells and sebum that cover it. This step is important to reduce the impedance between the electrodes and the scalp. Measure the head circumference to define the size of the electrode cap that will be used in the experiment.

Ask the subject to wear the electrode cap. Provide the instructions for a comfortable but correct positioning of the cap. Measure the distance between nasion and inion.

Likewise, measure the distance between the left and right pre-auricular points. Correct the position of the electrode cap. Put conductive gel in the electrode locations considered for the experiment.

The number of recording sites may vary as needed. Usually, we record from 64 scalp locations using a radio system. Place the recording electrodes on the correct locations.

Accompany the volunteer to the experimental room and ask the subject to sit in a comfortable position. Place external electrodes in periocular locations to record the electrooculogram. These signals will be used in the next steps, for correcting EEG artifacts induced by blinking and eye movements.

Turning the EEG acquisition system on and check the electrode impedance. Correct the impedance, as needed, according with the manufacturer's directions. Ask the subject to blink and move the eyes in different directions to ensure that the EOG is being correctly recorded by the external electrodes.

Adjust the location of the screen in the vertical direction, according with the view angle of the subject. Our screen consists of four light emitting diodes situated on the center of a 50x50 cm black screen, as vertexes of the square are 5x5 cm. Participants are seated approximately 70 centimeters from the screen so the are of the square of LEDs substends a visual angle of about four degrees.

Adjust the luminance level of the screen to the upper limit of the participants'comfortable level. Set the parameters of the visual stimulation. In our experiments, a continuous visual stimulation is presented where the light intensity is modulated at 10Hz.

Present the stimulus for the time required in the experiment. Pause the stimulation for two minutes. Pauses three times longer than the stimulation period are recommended.

Repeat the presentation steps 30 times. 30 runs of the experiment will ensure a high signal to noise ratio of the measurements. Nevertheless, a higher number of repetitions can be implemented in the experimental protocol.

Record the EEG using standard procedures. The experimental runs can be stored in a single file, or a different file can be created for each run. The next steps correspond to a standard EEG processing.

This processing is performed offline and can be modified as appropriate. Re-reference the recording using an average reference. Band pass filter the EEG signal, cut off frequencies can be modified as needed.

If necessary, convert the electrode coordinates to the international 10-20 System. Remove the ocular artifacts using appropriate procedures. To this end, different techniques can be used.

Segment the EEG data and epochs of appropriate length. Remove epochs containing EEG artifacts. Detrend the EEG epochs to direct current drifts.

Rearrange the epochs into a data matrix of N rows and M columns, in which N represents the number of recordings, and M, the number of epochs. Column wise, average the data set. To this end, the thirty epochs corresponding to the same time window in the different recordings need to be averaged in the time domain.

Compute the amplitude of the steady-state response at the end of the averaging using the Fast Fourier Transform. The amplitude of the steady-state response is defined as the spectral amplitude obtained at the frequency of the amplitude modulations of the sensory stimuli. Vector average the amplitude of an add hawk number of FFT bins at each side of the frequency of the response to calculate the residual noise level.

Plot the amplitude of the steady-state response and the RNL as a function of column index to explore the evolution of the parameters during the stimulation period.Results. Figure two illustrates changes in the waveform of the SSVEP resulting from the column wise averaging of epochs. Thirty recordings were obtained.

The neural oscillation time locked to the stimulation became evident as the column wise averaging was performed. Significantly, the period at which the SSVEP is generated can be observed in the traces corresponding with column one. In that column, 02 seconds of pre-stimulus baseline are plotted.

Therefore, the procedure described here allows to characterize not only the dynamics of the oscillatory response once the neural entrainment is already established, but also the engagement of the neural oscillations. The mean amplitude of the SSVEP decreased during the averaging of the first epochs of the columns and tended to stabilize afterwards. This behavior can be explained by the relatively high contribution of the noise to the response amplitude computed in the first averaged epochs which is attenuated as averaging is performed.

The standard deviation of the residual noise level remained relatively constant as the number of averaged epochs increased, which suggests that the recording conditions were stable along the experimental section. The results presented above determined the changes in the peak signal to noise ratio of the measurements. As averaging progressed, the peak signal to noise ratio increased as the number of average epochs increased up to 18, approximately.

Further increments in the number of averaged epochs did not significantly impact the quality of the signal. Finally, the dynamics of the steady-state visual evoked potential amplitude, and the residual noise level are represented in Fig. 4 Those dynamics were obtained by plotting the response parameters computed at the end of the column wise averaging of epochs as a function of the number of the columns as a function of time.

In this subject, the response amplitude gradually increased during the first 12 seconds, following stimulus onset. The time, which corresponds to the length of three epochs. As the stimulus persisted, the response consistently decreased during the following 12 seconds, and remained relatively constant afterwards.

These results cannot be explained by the behavior of the RNL, since this parameter was relatively constant during the stimulation period. The increase in the SSVEP amplitude following the stimulus onset, can be explained by integration processes, which result in the stabilization of the neural entrainment. The subsequent decrease in amplitude suggests the adaptation of the SSVEP to the sustained stimulation.

Nevertheless, these hypotheses need to be tested in controlled experiments with appropriated sample size. Computing the amplitude of the steady-state responses after the time domain averaging of independent runs implies analyzing only time-locked oscillations, those which survive the averaging. This procedure may filter relevant information regarding the dynamics of the response in individual trials.

However, it guarantees a sufficiently high signal to noise ratio. This aspect might be of particular significance when the responses are close to the electrophysiological threshold, a condition in which the detection of the entrainment can be compromised due to low signal-to-noise ratio of the measurement.