Abstract
Background Amplitude modulated transcranial alternating current stimulation (AM-tACS) has been recently proposed as a possible solution to overcome the pronounced stimulation artifact encountered when recording brain activity during tACS. In theory, AM-tACS does not entail power at its modulating frequency, thus avoiding the problem of spectral overlap between brain signal of interest and stimulation artifact. However, the current study demonstrates how weak non-linear transfer characteristics inherent in stimulation and recoring hardware can reintroduce spurious artifacts at the modulation frequency.
Method The input-output transfer functions (TFs) of different stimulation setups were measured. The setups included basic recordings of signal-generator and stimulator outputs as well as M/EEG phantom measurements. 6th-degree polynomial regression models were fitted to model the input-output TFs of each setup. The resulting TF models were applied to digitally generated AM-tACS signals to predict the location of spurious artifacts in the spectrum.
Results All four setups measured for the study exhibited low-frequency artifacts at the modulation frequency and its harmonics when recording AM-tACS. Fitted TF models showed nonlinear contributions significantly different from zero (all p < .05) and successfully predicted the frequency of artifacts observed in AM-signal recordings.
Conclusions Our results suggest that even weak non-linearities of stimulation and recording hardware can lead to spurious artifacts at the modulation frequency and its harmonics. Thus, findings emphasize the need for more linear stimulation devices for AM-tACS and careful analysis procedures taking into account these low-frequency artifacts.