From spinal central pattern generators to cortical network: integrated BCI for walking rehabilitation

Abstract

Success in locomotor rehabilitation programs can be improved with the use of brain-computer interfaces (BCIs). Although a wealth of research has demonstrated that locomotion is largely controlled by spinal mechanisms, the brain is of utmost importance in monitoring locomotor patterns and therefore contains information regarding central pattern generation functioning. In addition, there is also a tight coordination between the upper and lower limbs, which can also be useful in controlling locomotion. The current paper critically investigates different approaches that are applicable to this field: the use of electroencephalogram (EEG), upper limb electromyogram (EMG), or a hybrid of the two neurophysiological signals to control assistive exoskeletons used in locomotion based on programmable central pattern generators (PCPGs) or dynamic recurrent neural networks (DRNNs). Plantar surface tactile stimulation devices combined with virtual reality may provide the sensation of walking while in a supine position for use of training brain signals generated during locomotion. These methods may exploit mechanisms of brain plasticity and assist in the neurorehabilitation of gait in a variety of clinical conditions, including stroke, spinal trauma, multiple sclerosis, and cerebral palsy.

Figure 1

Research axes of MINDWALKER and BIOFACT projects. Parallel and convergent pathways using a dynamic recurrent neural network (DRNN, left part) and a central pattern generator (CPG, right part). The DRNN receives as input either the EMG signals from shoulder muscles mimicking the walking movement or the spontaneous EEG signals during walking. The CPG receives either steady-state somatosensory evoked potentials (SSSEPs), steady-state visual evoked potential (SSVEP), or classical P300 as starting signal. The SSSEPs are elicited by vibrotactile (tactors) stimulation on the foot sole mimicking walking patterns. A virtual reality stimulator (VRS) is used in order to generate an image of a walking mannequin to elicit SSVEP or to produce visual stimulation related to P300 speller.

Figure 2

Kinematics of the lower limb predicted by a DRNN receiving shoulder muscles EMG signals. (a) During the learning phase, smoothed and rectified EMG signals of the anterior and posterior deltoid muscle (AD, PD) are used as input while the elevation angles of the feet (FT), the shank (SK), and the thigh (TH) are the desired outputs. (b) Superimposition of the real and simulated elevation angle curves. (c) During the prediction phase, unlearned EMG used as input to the DRNN. (d) Superimposition of the real and simulated elevation angle curves produced by unlearned EMG data.

Figure 3

The PCPG is able to learn the frequency components of a periodic signal as well as the various phases and magnitudes. One major interest of PCPGs is the possibility to modify a learned pattern in amplitude or frequency in a smooth way. This figure is adapted from Righetti et al. [43].

Figure 4

PCPG performance as a function of the walking speed. (a) Superimposition of the real foot elevation angle (red) and the PCPG output (blue) by means of 7 oscillators. (b) The difference between SI values obtained with and without the interpolation is not significant. In this case, error bars are standard errors (modified from Duvinage et al. [46]).

Figure 5

Event-related spectral perturbation (ERSP) during walking cycle recorded in C3 and C4 electrodes for one subject. The stripped lines indicate the right heel strike event upon which the averaging was triggered. The power increase is represented in red color and the power decrease in blue color.

Figure 6

DRNN learning (a) and simulation (b) phases where two independent components (ICeeg_1-2) were used as input to the DRNN while the two principal components (PCk_1-2) of the 3 elevation angles foot, shank, and thigh of one lower limb kinematics were used as desired output (modified from [47]).

Figure 7

EEG activity during normal and obstacle steps performed on a treadmill. Grand averages of the initial EEG activity and topography (32 EEG electrodes placed on the scalp) from 12 subjects. The stance and swing phases were determined by the time period where all 12 subjects were on stance and swing (i.e., the individually shortest stance and swing phases, resp.). In normal steps a task-irrelevant and in obstacle steps a task-relevant acoustic signal was delivered at the onset of the right stance phase. Significant differences are indicated by asterisks (*P < 0.05, **P < 0.01) (modified with permission of Figure 3 from Haefeli et al. [11]).

Figure 8

Vibrotactile stimulation. (a) The time profile of the vibrotactile stimulation provided by heel and first metatarsal tactors is shown. The timing and the duration of vibration correspond to a gait cycle at 3 km/h speed [50]. (b) The evoked potential showing 3 main components (N100, P200, and P300) recorded at C2 electrode. The color maps show the distribution of these 3 evoked components. The C-2 Tactor is a linear actuator that has been optimized for use against the skin. The C-2 Tactor incorporates a moving “contactor” that is lightly preloaded against the skin. When an electrical signal is applied, the “contactor” oscillates perpendicular to the skin, while the surrounding skin area is “shielded” with a passive housing. Thus, unlike most vibrational transducers (such as common eccentric mass motors that simply shake the entire device), the C-2 provides a strong, point-like sensation that is easily felt and localized. For optimum vibrotactile efficiency, the C-2 is designed with a primary resonance in the 200–300 Hz range that coincides with peak sensitivity of the Pacinian corpuscle, the skin's mechanoreceptors that sense vibration. The subjects are seated with both feet on the floor, wearing a sandal on the left foot with a first tactor at the level of the heel and a second tactor at the level of the head of the first metatarsal. The stimulation of each tactor is made at a frequency of 300 Hz.