Real-Time Artifact Detection and Removal for Closed-Loop EEG-TMS

Abstract

Transcranial magnetic stimulation (TMS) combined with electroencephalography (EEG) is a non-invasive tool for studying brain connectivity and excitability. However, the EEG signals are often hindered by artifacts. Several signal-processing methods have been developed for correcting these artifacts offline. Yet, new promising EEG-TMS applications, such as closed-loop stimulation, would greatly benefit from artifact correction in real time. We present an algorithm for real-time attenuation of extracranial noise and removal of ocular artifacts from EEG-TMS data. Two established offline cleaning methods were implemented in a real-time setting: the source-estimate-utilizing noise-discarding (SOUND) algorithm and ocular-artifact removal with independent component analysis (ICA). This procedure cleans streamed raw data by multiplying every EEG sample with SOUND and ICA spatial filters, with a delay of less than 0.1 ms. The SOUND filter is constantly updated in a parallel process to react to changes in noise characteristics. In tests with pre-recorded EEG-TMS data, the proposed algorithm was fast enough for real-time use, removed ocular artifacts efficiently, and detected and cleaned contaminated channels automatically, leaving the noiseless channels intact. The algorithm can be used to detect and remove extracranial noise and ocular artifacts in real-time EEG and EEG-TMS experiments.