The Effect of Stimulation Intensity, Sampling Frequency, and Sample Synchronization in TMS-EEG on the TMS Pulse Artifact Amplitude and Duration

• Published in: IEEE transactions on neural systems and rehabilitation engineering Vol. 32; pp. 2612 - 2620
• Main Authors: Jamil, Zunaira, Saisanen, Laura, Demjan, Michal, Reijonen, Jusa, Julkunen, Petro
• Format: Journal Article
• Language: English
• Published: United States IEEE 2024
• Subjects: Adult; Algorithms; Electrodes; Electroencephalography; Electroencephalography - methods; Evoked Potentials, Motor - physiology; Female; Healthy Volunteers; Humans; Magnetic resonance imaging; Male; Motor Cortex - physiology; Motors; pulse artifact; Reliability engineering; Reproducibility of Results; sampling frequency; Silicon; stimulation intensity; Synchronization; TMS-EEG; Transcranial Magnetic Stimulation - methods; Young Adult
• ISSN: 1534-4320; 1558-0210; 1558-0210
• DOI: 10.1109/TNSRE.2024.3429176

Transcranial magnetic stimulation (TMS) coupled with electroencephalography (EEG) possesses diagnostic and therapeutic benefits. However, TMS provokes a large pulse artifact that momentarily obscures the cortical response, presenting a significant challenge for EEG data interpretation. We examined how stimulation intensity (SI), EEG sampling frequency (Fs) and synchronization of stimulation with EEG sampling influence the amplitude and duration of the pulse artifact. In eight healthy subjects, single-pulse TMS was administered to the primary motor cortex, due to its well-documented responsiveness to TMS. We applied two different SIs (90% and 120% of resting motor threshold, representing the commonly used subthreshold and suprathreshold levels) and Fs (conventional 5 kHz and high frequency 20 kHz) both with TMS synchronized with the EEG sampling and the conventional non-synchronized setting. Aside from removal of the DC-offset and epoching, no preprocessing was performed to the data. Using a random forest regression model, we identified that Fs had the largest impact on both the amplitude and duration of the pulse artifact, with median variable importance values of 1.444 and 1.327, respectively, followed by SI (0.964 and 1.083) and sampling synchronization (0.223 and 0.248). This indicated that Fs and SI are crucial for minimizing prediction error and thus play a pivotal role in accurately characterizing the pulse artifact. The results of this study enable focusing some of the study design parameters to minimize TMS pulse artifact, which is essential for both enhancing the reliability of clinical TMS-EEG applications and improving the overall integrity and interpretability of TMS-EEG data.