Detection of self-paced reaching movement intention from EEG signals

Abstract

Future neuroprosthetic devices, in particular upper limb, will require decoding and executing not only the user's intended movement type, but also when the user intends to execute the movement. This work investigates the potential use of brain signals recorded non-invasively for detecting the time before a self-paced reaching movement is initiated which could contribute to the design of practical upper limb neuroprosthetics. In particular, we show the detection of self-paced reaching movement intention in single trials using the readiness potential, an electroencephalography (EEG) slow cortical potential (SCP) computed in a narrow frequency range (0.1-1 Hz). Our experiments with 12 human volunteers, two of them stroke subjects, yield high detection rates prior to the movement onset and low detection rates during the non-movement intention period. With the proposed approach, movement intention was detected around 500 ms before actual onset, which clearly matches previous literature on readiness potentials. Interestingly, the result obtained with one of the stroke subjects is coherent with those achieved in healthy subjects, with single-trial performance of up to 92% for the paretic arm. These results suggest that, apart from contributing to our understanding of voluntary motor control for designing more advanced neuroprostheses, our work could also have a direct impact on advancing robot-assisted neurorehabilitation.

Keywords: BCI; EEG; rehabilitation; self-paced protocol; stroke; voluntary movements.

Figure 1

Experimental setup for Experiment 1(left) and Experiment 2 (right).

Figure 2

The timeline of the experimental protocol. Each trial starts when the subject places their hand on the center button. Next, the auditory cue informs the subject which direction to reach. After a delayed period of more than 2 s, he releases his hands from the home position and reaches towards the target. In order to complete the movement, the subject returns back to the home position before starting the next trial. Only center-out reaching periods are considered. The average T_onset across all subjects in Experiment 1 is 5.03 ± 1.77 s.

Figure 3

Regression coefficients of EOG components plotted on a topographical map, showing the effect of eye movement on scalp electrodes using signal re-referenced with all 64-channels recorded from one of the subjects. The rightmost figure shows the sum of the contributions of both vertical EOG and horizontal EOG.

Figure 4

(Top) The weights of EOG artifacts by re-referencing the signals with 41-channels and (bottom) 34-channels. This figure shows the EOG coefficients from the calibration session of one of the subjects participating in the experiments.

Figure 5

Grand averages of SCPs for all the right-handed subjects participating in Experiment 1. EEG signals are filtered between 0.1 and 1 Hz. t = 0 corresponds to the movement onset.

Figure 6

Grand averages of SCPs, filtered between 0.1 and 1 Hz, for the paretic arm of stroke patient lg (right arm) from Experiment 2. t = 0 corresponds to the movement onset.

Figure 7

Each topoplot shows the normalized discriminant power index of each channel for a single healthy subject in Experiment 1.

Figure 8

The topographic maps show the normalized discriminant power index of each channel for the left and right hand for the control subjects (cg and gc) and the stroke subjects (dpm and lg). Plots highlighted with a blue frame refers to the paretic arm of the patients.

Figure 9

Selected EEG samples to build the training set of the movement intention classifier.

Figure 10

Single trial performances of movement intention detection for all subjects in Experiment 1 using SCPs in the frequency range [0.1–1] Hz during the time interval (−2, 1) s with respect to the actual movement onset. Y-axis of the plots represents the movement intention detection rate. The magenta line depicts the onset from arm muscular activation (−263 ± 40 ms on average across all subjects). The green line depicts the first time a group of five consecutive samples has a TPR significantly above chance level (p < 0.05), which is shown as a red line. The gray and red shaded regions bounding the performance curves indicates their standard deviation at each point. Note that the variance of the random performance is so small that the red shaded area is barely visible.

Figure 11

Time of movement intention detection comparison between pre-selected channels set and best selected channel using CVA techniques.

Figure 12

Single trial performances of movement intention detection for all subjects in Experiment 2 (both left and right arm reaching movement) using SCPs in the frequency range [0.1–1] Hz during the time interval (−2, 1) s with respect to the actual movement onset. This figure has a similar format to Figure 10. Plots highlighted with a blue frame refers to the paretic arm of the stroke subjects.

Figure 13

Comparison of detected movement intention when TPR is above chance level (p < 0.05) between using pre-selected channel set and best selected channel from the data using CVA technique.

Figure 14

Single trial detection of movement intention from EMG activity for all subjects in Experiment 1. Y-axis of the plots represents movement intention detection rate.

Figure 15

Single trial detection of movement intention from EMG activity for all subjects and hands in Experiment 2. Plots highlighted with a blue frame refers to the paretic arm of the patients. Y-axis of the plots represents movement intention detection rate.

Figure 16

Detection of movement intention during the non-movement intention period for all subjects in Experiment 1. Time 0 s refers to the delivery of the auditory target cue. Y-axis of the plots represents movement intention detection rate.

Figure 17

Detection of movement intention during the non-movement intention period for all subjects and hands in Experiment 2. Time 0 s refers to the delivery of the visual target cue. Plots highlighted with a blue frame refers to the paretic arm of the patients. Y-axis of the plots represents movement intention detection rate.

Figure 18

Each pixel refers to the single trial performance of movement intention detection for all subjects in Experiment 1 using signals filtered in various frequency ranges (Y-axis). The dotted line in magenta refers to the EMG activation for each subject.