Combined transcranial magnetic stimulation (TMS) and electroencephalography (EEG) often suffers from large muscle artifacts. Muscle artifacts can be removed using signal-space projection (SSP), but this can make the visual interpretation of the remaining EEG data difficult. We suggest to use an additional step after SSP that we call source-informed reconstruction (SIR). SSP–SIR improves substantially the signal quality of artifactual TMS–EEG data, causing minimal distortion in the neuronal signal components.
In the SSP–SIR approach, we first project out the muscle artifact using SSP. Utilizing an anatomical model and the remaining signal, we estimate an equivalent source distribution in the brain. Finally, we map the obtained source estimate onto the original signal space, again using anatomical information. This approach restores the neuronal signals in the sensor space and interpolates EEG traces onto the completely rejected channels.
The introduced algorithm efficiently suppresses TMS-related muscle artifacts in EEG while retaining well the neuronal EEG topographies and signals. With the presented method, we can remove muscle artifacts from TMS–EEG data and recover the underlying brain responses without compromising the readability of the signals of interest.