Probing cortical excitability using rapid frequency tagging

Abstract

Frequency tagging has been widely used to study the role of visual selective attention. Presenting a visual stimulus flickering at a specific frequency generates so-called steady-state visually evoked responses. However, frequency tagging is mostly done at lower frequencies (<30 Hz). This produces a visible flicker, potentially interfering with both perception and neuronal oscillations in the theta, alpha and beta band. To overcome these problems, we used a newly developed projector with a 1440 Hz refresh rate allowing for frequency tagging at higher frequencies. We asked participants to perform a cued spatial attention task in which imperative pictorial stimuli were presented at 63 Hz or 78 Hz while measuring whole-head magnetoencephalography (MEG). We found posterior sensors to show a strong response at the tagged frequency. Importantly, this response was enhanced by spatial attention. Furthermore, we reproduced the typical modulations of alpha band oscillations, i.e., decrease in the alpha power contralateral to the attentional cue. The decrease in alpha power and increase in frequency tagged signal with attention correlated over subjects. We hereby provide proof-of-principle for the use of high-frequency tagging to study sensory processing and neuronal excitability associated with attention.

Fig. 1

Schematic representation of the experimental paradigm. After an attentional cue, a house-face pair was presented at 63 and 78 Hz (counterbalanced over trials). In 20% of the trials, one of the images was flipped vertically and required participant's response. In 5% of the trials (catch trials), the flip was in the hemifield opposite to the cued side and participants had to ignore this event.

Fig. 2

Event-related fields for a representative participant showed clear responses at the tagging frequencies. Note that the frequency tagged signals were presented with the same phase over trials. (A) Broadband (black line) and narrowband (red line) trial-averaged ERFs for 63 Hz stimulus (presented right) for the left occipital sensors (see panel C). (B) Trial-averaged ERFs for 78 Hz stimulus (presented right) for the left occipital sensors. (C) Left and right occipital MEG sensors that covered areas with the stronger power at the tagging frequencies for all the participants were used in the analysis. (D) Normalized group-level power spectra for the left sensors when the tagged image was presented at 63 Hz and 78 Hz in the right hemifield. Prior to computing individual power spectra, the trials were normalized by the standard deviation of time series over sensors. The line noise with peak near 50 Hz was cut out in the plot.

Fig. 3

Attention modulates power in the alpha band and at the tagging frequencies. (A) Time-frequency representation of the attention modulation index (AMI). The AMI reflects the power modulation in the sensors contra-versus ipsilateral to the attended hemifield for combined left and right occipital sensors (see Fig. 2C for sensors selection). The power was calculated per trial and then averaged. Black line indicates onset of the frequency tagged stimuli; the cue onset was at −0.5 s. (B) The AMI (averaged over time bins 0.5–1.5 s) at the group level. Dashed lines indicate p-values of the t-test comparing modulation index against zero (over participants). The effect is highly robust in the 8–12 Hz alpha band and at 63 and 78 Hz even if multiple comparisons over frequencies are considered. (C) Relative power change compared to the baseline (−1, −0.5 s) at the left sensors for trials “attention left” (cyan line; ipsilateral to the cue) and “attention right” (blue line; contralateral to the cue). (D) The same as (C) but for the right sensors for trials “attention right” (orange line; ipsilateral to the cue) and “attention left” (red line; contralateral to the cue).

Fig. 4

Group average topography maps of the AMI in the alpha band (10 ± 2 Hz) and tagging frequencies (63 and 78 Hz). Black dots indicate MEG sensors at which amplitude modulation index was significantly different from zero (p < 0.05, cluster-based permutation).

Fig. 5

Relationship between the modulation of alpha power and frequency tagging. (A) Spatial masks for the alpha and tagging frequencies. The masks were obtained by selecting sensors expression the 10% of largest absolute AMI values. (B) Scatter plot of individual AMI relating the alpha power modulation and the power combined for the tagging frequencies. Subjects with a strong alpha power modulation with attention were also subjects with a strong modulation of the tagged signals.