Projecting out muscle artifacts from TMS-evoked EEG

Abstract

Transcranial magnetic stimulation combined with electroencephalography is a powerful tool for probing cortical excitability and connectivity; we can perturb one brain area and study the reactions at the stimulated and interconnected sites. When stimulating areas near cranial muscles, their activation produces a large artifact in the electroencephalographic signal, lasting tens of milliseconds and masking the early brain signals. We present an artifact removal method based on projecting out the topographic patterns of the muscle activity. Although the brain and muscle components overlap both temporally and spectrally, the fact that muscle activity is present also at frequencies higher than 100 Hz, while brain signal is mostly restricted to frequencies lower than that, allows us to study the high-frequency muscle activity without brain contribution. We determined the muscle activity topographies from data highpass-filtered at a 100-Hz cutoff frequency using principal component analysis. Projecting out the topographies of the principal components which explain most of the variance of the high-frequency data reduces not only the high-frequency activity but also the low-frequency muscle contribution, because the topography produced by a muscle source can be expected to be the same regardless of the frequency. The method greatly reduced the muscle artifact evoked by stimulation of Broca's area, while a significant brain signal contribution remained. Improvement in the signal-to-artifact ratio, defined as the relative amplitudes of brain signals peaking after 50 ms and the first artifact deflection, was of the order of 10-100 depending on the number of projections. The presented artifact removal method enables one to study the cortical state when stimulating areas near the cranial muscles.