Poster Z12 - Variability in Motor Related ERPs During Upper Limb Reaching Tasks

Introduction: Brain computer interface (BCI) technology has the potential to restore motor capabilities to individuals who are paralyzed. Presently, intracortical BCIs have the power to translate populations of neural action potentials into motor commands to actuate 7-degree-of-freedom robotic arms [1]. However, the cost and safety risks posed by surgically implanting the electrode array and acquisition hardware are often prohibitive. To work towards our long-term goal of developing an affordable, safe, convenient, and reliable BCI that controls the motion of prosthetic devices, we are analyzing surface EEG (sEEG) features that could be used as predictors of movement direction in an arm reaching task. Conventional event-related potential (ERP) analysis has focused on a visual or auditory response to a stimulus cue in the posterior visual or auditory processing centers of the brain. In BCI research, visual cues are used to allow users to select different locations on a screen. In this work, we explore the possibility of a motor-related ERP in the frontal lobe, where the motor and pre-motor cortex are located. We determined the differences in ERPs across 8 different reach directions and examined variability across directions in comparison with variability for the same direction across subjects.

Materials and

Methods: sEEG was acquired from nine subjects performing a center-out reaching task with an affordable MP-160 amplifier system (Biopac, Goleta, CA) at 500 samples/s with 5000x gain, and 2-pole bandpass filtering from 0.5 to 35 Hz at frontal and central electrodes on each hemisphere and an additional 2-pole Butterworth 60-Hz notch filter.  Eyeblink artifact reduction was achieved by performing linear regression between the electrooculogram (EOG) channel and each EEG channel in custom code written in MATLAB (Mathworks, Inc. Natick, MA). These linear regression fits were then applied to the EOG channel and subtracted from each EEG channel. Event-related potentials (ERPs) were also computed using custom MATLAB code. All methods were approved by California State University, Los Angeles IRB, under protocol #1898091-1.

Results, Conclusions, and Discussions: Comparisons between ERPs for each of the 8 directions revealed differences, specifically in polarity of the main peaks, as well as peak time and amplitude, across the eight reach directions, but there is significant variability across trials. This decrease in variability was more prominent on some channels for certain peaks than others. For example, the decrease in variability, or increase in consistency, was prominent for the N1 (the first negative peak) peak time, as the standard deviation for a given direction across subjects was 43% lower than the standard deviation across directions averaged across subjects. The ratio of standard deviation across directions for a given subject to standard deviation across subjects for a given direction was calculated and, for N1, found to be 0.57. Exemplary ERPs illustrating the consistency for a given direction across subjects and discriminability across directions are shown in Fig. 1.

Our results show that motor actions, such as arm reaches, appear to be encoded in event related potential features, such as peak time and peak width, and indicate that upper limb reach direction could be discriminated from surface EEG features. Our low-cost sEEG system was very susceptible to noise despite the use of the central ground electrode and mastoid references, as well as the shielding of cables. Occasionally, the baseline signal before t = 0 (cue onset) was still unsettled, and part of our future work will entail addressing motion artifact and EMI noise issues. Nevertheless, the ERP amplitudes were still more than an order of magnitude larger than this baseline noise. The results of this study indicate the discriminability of surface EEG-based event-related potentials in the frontal regions across eight directions of upper limb reach. Such results suggest the promise of developing non-invasive BCIs for reproducing or controlling upper limb motor tasks.

Acknowledgements (Optional):