Can Electroencephalogram (EEG) Signals Predict Postural Balance Performance?

Abstract

Although it is argued that EEG signals lack sufficient signal-to-noise ratio, bandwidth, and information content to decode body segment kinematics (Lebedev and Nicolelis, 2006), this pilot study examined the feasibility of decoding lower extremity postural balance kinematics (i.e., anterior-posterior [A-P] and medial-lateral [M-L] deviation) utilizing EEG signals with reasonable accuracy. The study utilized single trial data of 5 participants that completed a postural balance task on a Biodex balance system when balance platform was set at a relatively unstable level. The neural response during the balance task was collected for 19 sites on the scalp. The time series data of 26 cortical locations were calculated. Following velocity calculation, resampling, and smoothing of the data, a linear decoding method was followed for both scalp and cortex data that utilized a fixed lag of 100 milliseconds and 10×10 fold cross validation procedure. Decoding accuracy was measured by estimating the correlation coefficient between measured and model reconstructed velocities. The average correlation coefficient (standard deviation in parenthesis) using scalp EEG signals was 0.356 (0.218) and 0.41 (0.179) for A-P and M-L velocities, respectively. The average correlation using localized cortical time series was 0.396 (0.238) and 0.458 (0.175) for A-P and M-L velocities, respectively. The findings were argued to be reasonably promising compared to other recent findings.