Frequency Tagging

Scalp electrophysiological recordings (EEG, MEG) are characterised by multiple fluctuations spanning all time scales. While we use EEG and MEG to track the neural responses to a presented stimulus, most of this activity is endogenous, i.e. not related to our experimental stimulation. Relating a specific aspect of a stimulus to its specific brain response might therefore be problematic if the response is weak, if it overlap with the response to another aspect of the same stimulus, or more simply if the stimulus presentation time is limited.

However, if the temporal course of the stimulus presentation is systematically periodic, one may assume that the neural population coding for that stimulus will also oscillate with the same period. The frequency of the stimulus thus provides a frequency 'tag' to identify the associated brain response. This response will be easily dissociated from the broadband endogenous activity because it will emerge as a peak in the power spectrum of the signal, resulting in a much higher signal-to-noise ratio (SNR) than the one obtained with the classical ERP analysis.

One example of the efficiency of frequency-tagging compared to ERP analysis is illustrated in the following figure (adapted from Buiatti, 2009) showing the EEG activity recorded at one occipital electrode, related to the visual presentation of a black&white checkerboard shifting colors every 250 ms. Due to the wide fluctuations of the stimulus-unrelated ongoing brain activity, no response is visible at the single trial level (top-left panel, t=0 indicates the time of color reversal). By averaging over 40 trials, the overlapped ongoing activity slowly cancels out and we obtain a clear event-related potential with an occipital topography (top-righ panel). However, we can exploit the precise periodicity of the stimulation at 4 Hz (bottom-left panel) to extract the stimulus-related response in the frequency-domain (bottom-right panel).  Since the ongoing activity does not specifically interfere with the frequency peak at 4 Hz, the SNR obtained with frequency-tagging is 5 times higher than the one obtained with the ERP. 

In collaboration with cognitive neuroscientists, I have used frequency-tagging paradigms to investigate the neural underpinnings of several perceptual and cognitive processes both in adults (Buiatti et al., 2009, Forget et al., 2010) and during development (Kabdebon et al., 2015, Buiatti et al., 2019) (see the Cognitive Neuroscience section for details).

More recently, in collaboration with students and researchers with bioengineering training, I am exploring the potentiality of frequency-tagging for extracting a reliable neural response in challenging contexts like developmental/pathological populations in out-of-the-lab settings. More specifically, we are investigating what is the optimal method to measure neural responses with frequency-tagging paradigms in the limit of very short data duration and with low stimulation frequencies (necessary for targeting higher-order perceptual processes) with traditional high-density EEG (Saretta & Buiatti, in prep.) and with low-density wearable EEG (Kartsch et al., 2022). Both studies show that by using the appropriate measures, even for very low-frequency stimulations (1 Hz) it is possible to retrieve a reliable response in as short as 10 seconds of clean data recordings.

References:

Buiatti M, 

Analisi multidimensionale della dinamica neurale di un processo cognitivo, 

in Bioingegneria per le Neuroscienze Cognitive, Edizioni Patron, Bologna (Italy), (2009). Book chapter, in Italian.

Buiatti M, Pena M, Dehaene-Lambertz G,

Investigating the neural correlates of continuous speech computation with frequency-tagged neuroelectric responses,

Neuroimage 44, 509-519 (2009).

Forget J, Buiatti M, Dehaene S,

Temporal integration in visual word recognition,

Journal of Cognitive Neuroscience 22(5), 1054-1068 (2010).

Kabdebon C, Pena M, Buiatti M, Dehaene-Lambertz G,

Electrophysiological evidence of statistical learning of long-distance dependencies in 8-month-old preterm and full-term infants,

Brain and Language 148, 25-36 (2015)

Buiatti M, Di Giorgio E, Piazza M, Polloni C, Menna G, Taddei F, Baldo E, Vallortigara G,

Cortical route for facelike pattern processing in human newborns, 

Proceedings of the National Academy of Sciences, (2019). 

Kartsch V, Kumaravel VP, Benatti S, Vallortigara G, Benini L, Farella E, Buiatti M,

Efficient Low-Frequency SSVEP detection with wearable EEG using Normalized Canonical Correlation Analysis.

Sensors 22(24), 9803 (2022)

Buiatti M, Saretta D

The power of power: Estimating steady-state visual evoked potentials in the limit of short data duration and low stimulation frequency.

In preparation.