μ-Rhythm Extracted With Personalized EEG Filters Correlates With Corticospinal Excitability in Real-Time Phase-Triggered EEG-TMS

Abstract

Ongoing brain activity has been implicated in the modulation of cortical excitability. The combination of electroencephalography (EEG) and transcranial magnetic stimulation (TMS) in a real-time triggered setup is a novel method for testing hypotheses about the relationship between spontaneous neuronal oscillations, cortical excitability, and synaptic plasticity. For this method, a reliable real-time extraction of the neuronal signal of interest from scalp EEG with high signal-to-noise ratio (SNR) is of crucial importance. Here we compare individually tailored spatial filters as computed by spatial-spectral decomposition (SSD), which maximizes SNR in a frequency band of interest, against established local C3-centered Laplacian filters for the extraction of the sensorimotor μ-rhythm. Single-pulse TMS over the left primary motor cortex was synchronized with the surface positive or negative peak of the respective extracted signal, and motor evoked potentials (MEP) were recorded with electromyography (EMG) of a contralateral hand muscle. Both extraction methods led to a comparable degree of MEP amplitude modulation by phase of the sensorimotor μ-rhythm at the time of stimulation. This could be relevant for targeting other brain regions with no working benchmark such as the local C3-centered Laplacian filter, as sufficient SNR is an important prerequisite for reliable real-time single-trial detection of EEG features.

Keywords: EEG-TMS; brain-state-dependent stimulation; corticospinal excitability; sensorimotor oscillations; spatial filtering.

Figure 1

Methods and results. (A) EEG cap layout. Channels used for estimation of the individual spatial filters are marked with circles. Red circles indicate the channels used for determination of individual μ-peak-frequency. (B) Example 1/f-corrected spectrum used for determination of individual μ-peak frequency. Frequency bands used for the computation of SSD filters. Marked in yellow are the individual μ-peak frequency ±2 Hz, in gray the flanking noise frequency bands. (C) Example SSD spatial filter (right) computed from resting state EEG activity and the standard C3-centered Laplacian filter (left). (D) Group median MEP amplitudes for the respective filters, normalized by global median. p-values for Wilcoxon signed-rank test, multiple comparison corrected for the two types of filters, N = 15. (E) Modulation of MEP amplitudes by μ-phase as assessed by the N/P fraction for the respective filters, Laplace N/P-fraction vs. SSD N/P-fraction with 2.5–97.5th-percentile confidence intervals for each subject, N = 15.

Figure 2

Illustration of phase shifts in sensor space. (A) Topography with three neighboring channels (FCC3h, C3, CCP5h) selected as center electrodes for the local spatial filter. (B) Resting-state EEG sensor space signals of one participant spatially filtered by a Laplacian filter centered on the selected electrode. The events are aligned to the troughs of the C3-centered Laplacian signal. A systematic phase shift is visible in the FCC3h- and CCP5h-centered signal respective to the C3-centered Laplacian signal troughs, with 58.5° and −29.7°, respectively.