Evaluation of artifact reduction methodologies in TMS-EEG signals

Abstract

Evoked-Potentials generated by Transcranial Magnetic Stimulation (TMS) are essential for assessing cortical excitability and effective connectivity in clinical research. However, obtaining artifact-free TMS-Evoked Potentials (TEPs) is challenging due to contamination of the electroencephalographic (EEG) signal by artifacts generated during the administration of the TMS-pulse. Various pre-processing approaches have been developed to remove these artifacts, although the degree to which these methods reduce artifacts or distort the signal is not yet fully understood. This study aimed to evaluate the variability in TEP waveforms resulting from the use of different pre-processing approaches. For this purpose, we used TMS-EEG data from six subjects obtained from stimulation of the left dorsolateral prefrontal cortex (DLPFC), employing five different techniques and comparing the results. Differences in TEP amplitudes are observed but they did not reach statistical significance. The TEP topographies showed correlations between methodologies for latencies above 100 ms, while waveforms at latencies below this value showed less similarity. In addition, the ARTIST technique is discarded due to the amplitude values obtained. Altogether, these findings suggest that the BIOART and SOUND-SSP-SIR techniques generate signals more similar to the ideal ground truth of TEPs and to the activation and inhibition patterns of each peak. Although the choice of pre-processing approach influences our final TEP result, this impact does not appear to be critical. However, more studies are needed to understand the advantages and disadvantages of the various methods and to establish the ground truth of the clean signal in TMS-EEG.