The impact of Independent Component Analysis on TMS-evoked potentials: a within-subject comparison across motor and prefrontal areas

Abstract

Background: Independent component analysis (ICA) is a common method to remove artifacts and improve signal quality in transcranial magnetic stimulation combined with electroencephalography (TMS-EEG) data. However, the impact of applying different rounds of ICA for TMS-EEG datasets are lacking. Here, we investigated the impact of applying zero, one, or two rounds of ICA on TMS-evoked potentials (TEPs) in the motor cortex and prefrontal cortex using a within-subject design. Methods: Twenty-three healthy participants received 51 single-pulses over the left primary motor cortex (M1) and dorsolateral prefrontal cortex (DLPFC), with simultaneous EEG-recording. Data were preprocessed three times: with zero, one, or two rounds of ICA. We compared the TEP-amplitudes and local mean field potential-area under the curve (LMFP-AUC). Results: After M1 stimulation, preprocessing two ICA rounds resulted in significantly more N40-peaks identified, and smaller P30 (3.84 µV ± 0.78) and P60 (3.08 µV ± 0.96) amplitudes compared to 0 ICA rounds. After DLPFC stimulation, zero ICA rounds led to significantly fewer identified P30 (n = 4) and N40-peaks (n = 5), and larger P60 (5.63 µV ± 1.34) and P180-amplitudes (8.62 µV ± 0.76) compared to 0 ICA. The N100-component remained stable across ICA conditions for both brain areas. Discussion: Our results showed a large impact of ICA, compared to zero ICA rounds, particularly after DLPFC stimulation; likely due to increased eye-blink artifacts. We therefore recommend implementing eye-blink artefact removal during the first round of ICA. In M1, the addition of an ICA round only impacted the early-TEPs. This study is the first to directly compare ICA-effects on both motor and non-motor TMS-EEG data using a within-subject design.