Background: Deep brain stimulation (DBS) has treatment efficacy in neurological and psychiatric disorders such as Parkinson's disease and major depression. Electroencephalography (EEG) is a versatile neurophysiological tool that can be used to better understand DBS treatment mechanisms. DBS causes artifacts in EEG recordings that preclude meaningful neurophysiological activity from being quantified during stimulation.

New method: In this study, we modeled the DBS stimulation artifact and illustrated a technique for removing the artifact using matched filters. The approach was validated using a synthetically generated DBS artifact superimposed on EEG data. Mean squared error (MSE) between the recovered signal and the artifact-free signal was used to quantify the effectiveness of the approach.

Results: The DBS artifact was characterized by a series of narrow band components at the harmonic frequencies of DBS stimulation. The filtering approach successfully removed the DBS artifact with MSE value being less than 2% of base signal power for the typical stimulation and recording setups. General guidelines on how to setup DBS EEG studies and configure the subsequent artifact removal process are described.

Comparison with previous method: To avoid stimulus artifacts, a number of EEG studies with DBS subjects have resorted to turning the stimulator off during recording, while other studies have used low pass filters to remove artifacts and look at frequencies well below 50 Hz.

Conclusions: This study establishes a method through which DBS artifact in EEG recordings can be reliably eliminated, thereby preserving a meaningful neurophysiological signal through which to better understand DBS treatment mechanisms.