Estimating steady-state evoked potentials in the limit of short data duration and low stimulation frequency

Abstract

AbstractBecause of their high signal-to-noise ratio and robustness to artifacts, Steady-State Evoked Potentials (SSEP) - the periodic responses elicited by periodic stimulation designs - are increasingly used in human neuroscience for measuring stimulus-specific brain responses in a short presentation time. While widely applied to measure sensory responses with stimulation frequencies higher than 8 Hz, they are also successful to investigate high-order processes and/or early development characterized by slower time scales, requiring very low stimulation frequencies around 1 Hz. However, applications of these low frequency paradigms on developmental or clinical populations, typically relying on very short data recordings, pose a methodological challenge for SSEP estimation. Here we tackled this challenge by investigating the method of analysis that most efficiently compute SSEP at low stimulation frequencies in the limit of short data, and by estimating the minimum data length necessary to obtain a reliable response. We compared the performance of the three most commonly used measures of SSEP (power spectrum (PS), evoked power spectrum (EPS) and inter-trial coherence (ITC)) at progressively shorter data segments both on simulated data and on EEG responses to on-off checkerboard stimulation at two ‘low’ frequencies (4 Hz and 0.8 Hz). Results, consistent between simulated and real data, show that while for long data length EPS and ITC outperform PS, for short data length the three measures are equivalent, and the crucial parameter is the length of the sliding window over which each measure is computed: the longer the better for PS and EPS, whereas the opposite occurs for ITC. For the analysed dataset, the shortest data length required to estimate a reliable SSEP is as short as 8 cycles of stimulation, independently from the stimulation frequency. This study provides practical indications for reliable and efficient application of low-frequency SSEP designs on short data recordings.