Induction of Neural Plasticity Using a Low-Cost Open Source Brain-Computer Interface and a 3D-Printed Wrist Exoskeleton

Abstract

Brain-computer interfaces (BCIs) have been proven to be useful for stroke rehabilitation, but there are a number of factors that impede the use of this technology in rehabilitation clinics and in home-use, the major factors including the usability and costs of the BCI system. The aims of this study were to develop a cheap 3D-printed wrist exoskeleton that can be controlled by a cheap open source BCI (OpenViBE), and to determine if training with such a setup could induce neural plasticity. Eleven healthy volunteers imagined wrist extensions, which were detected from single-trial electroencephalography (EEG), and in response to this, the wrist exoskeleton replicated the intended movement. Motor-evoked potentials (MEPs) elicited using transcranial magnetic stimulation were measured before, immediately after, and 30 min after BCI training with the exoskeleton. The BCI system had a true positive rate of 86 ± 12% with 1.20 ± 0.57 false detections per minute. Compared to the measurement before the BCI training, the MEPs increased by 35 ± 60% immediately after and 67 ± 60% 30 min after the BCI training. There was no association between the BCI performance and the induction of plasticity. In conclusion, it is possible to detect imaginary movements using an open-source BCI setup and control a cheap 3D-printed exoskeleton that when combined with the BCI can induce neural plasticity. These findings may promote the availability of BCI technology for rehabilitation clinics and home-use. However, the usability must be improved, and further tests are needed with stroke patients.

Keywords: brain–computer interface; exoskeleton; motor imagination; neural plasticity; neurorehabilitation.

Figure 1

Timeline of the experiment; the approximate duration of each block is indicated in parentheses. First, the subject was familiarized with transcranial magnetic stimulation (TMS) and motor imagination (MI), and the EEG cap was mounted. Next, the brain-computer interface (BCI) was calibrated, followed by the identification of the optimal stimulation site (hotspot) and intensity (RTh). The pre-intervention TMS, post-intervention TMS, and post-30 min intervention TMS were identical. After the pre-intervention TMS, the threshold for each subject was tested with an online BCI and changed if needed. Afterwards, the intervention started, and it was stopped when the subject reached 50 correct pairings between motor imagination (MI) and movement of the exoskeleton. The post-30 min intervention TMS started 30 min after the BCI intervention ended.

Figure 2

Motor-evoked potential (MEP) from a representative subject (post-intervention transcranial magnetic stimulation measurement for subject 1). The peak around 25 milliseconds is the stimulation artefact from the transcranial magnetic stimulation.

Figure 3

Overview of the hardware setup. The Arduino and Linear Actuator Control board were mounted on the exoskeleton. The EEG electrodes were connected through wires to the Open BCI board from which the signals were transmitted through wireless communication to the PC running OpenViBE. Once an imagined wrist extension was detected a trigger was sent through wireless communication to the Arduino on the exoskeleton. The Arduino was connected to the Linear Actuator Control board with a wire. The Linear Actuator Control board was powered with a 12 V power supply. The motor was connected to the Linear Actuator Control board with a wire.

Figure 4

View of the 3D-printed exoskeleton. The illustration is not drawn to scale. The surfaces that were in contact with the forearm and hand were padded with foam to improve the comfort. The exoskeleton was fixated to the subject’s hand and forearm with Velcro straps (A). ‘LAC’: Linear Actuator Control.

Figure 5

Summary of the MEP results. (a) Averaged MEP amplitudes (in mV) across the subjects, the vertical black line represents the standard deviation across subjects. The MEPs from the measurement 30 min after the intervention (Post 30) were significantly higher (denoted by *) than those from the measurement before the intervention (Pre). (b) MEP changes (in percent) from the measurement before the intervention to the measurement immediately after the intervention (Pre-Post) and 30 min after the intervention (Pre-Post 30).