Brain Computer Interfaces

Brain-computer interface (BCI) technology utilizes brain signals instead of spinal or peripheral motor systems to drive external devices (Birbaumer et al. 2006; Collinger et al. 2013). These devices act as assistive technology to help individuals with SCI complete activities of daily living, without requiring physical movement (Huggins et al. 2015). In order to control a BCI, the user’s brain activity is recorded (via a neural recording device, e.g. EEG) while performing or thinking of performing a motor movement (Collinger et al. 2013; Van Dokkum et al. 2015). After recording brain activity, the information is decoded and turned into visual, auditory or haptic feedback and even the control of external devices to help facilitate movement (Collinger et al. 2013; Van Dokkum et al. 2015). Besides helping to facilitate movement, BCI technology may promote neuroplasticity through the recruitment of brain areas involved in motor planning and execution to operate training devices (Van Dokkum et al. 2015). However, BCI technology has only recently emerged a rehabilitative treatment following SCI, therefore, the evidence base for this intervention rather limited.

The methodological details and results of eight studies evaluating BCI for upper extremity rehabilitation in SCI patients are presented in Table 8.

Author Year


Research Design

PEDro Score

Sample Size

Methods Outcomes
Osuagwu et al., 2016





Population: Mean age=51.7±18.4 yr; Gender: males=12; Time since injury: Not reported; Level of injury: C4 – C7; Severity of injury: AISA A=0, B=4, C=8, D=0.

Intervention: Participants were randomized to receive 20 sessions of BCI controlled FES (n=7) or the same number of sessions of passive FES (n=5), on both hands. Outcome measures were assessed at baseline and following treatment.

Outcome Measures: Event related desynchronization (ERD); Somatosensory evoked potential (SSEP); ROM; MMT.

1.     Patients in both groups initially had intense ERD during movement that was not restricted to the sensory-motor cortex.

2.     Following treatment, ERD cortical activity restored towards the activity in able bodied individuals in the BCI-FES group only.

3.     SSEP returned in 3 patients in the BCI-FES group, while no significant changes were observed in the FES alone group (p>0.05).

4.     All patients demonstrated increased ROM (median ROM for flexion and extension = 9.9 to 25.2) in both wrists following therapy except for one participant.

5.     MMT significantly improved in all muscles groups in the BCI-FES group (p<0.05), while the FES group improved shoulder muscles or muscles involved in controlling flexion (p<0.05).

Athanasiou et al., 2017




Population: SCI: Mean age=46.0±17.6 yr; Gender: males=8, females=2; Level of injury: T4 – C8; Severity of injury: AISA A=1, B=2, C=1, D=6.

Control: Mean age=46.2±18.2 yr; Gender: males=8, females=2.

Intervention: Participants with (n=10) or without SCI (n=10) operated two robotic arms via wireless commercial BCI, using motor imagery to perform 32 different upper extremity movements. Outcome measures were assessed after five training sessions with the BMI.

Outcome Measures: Training skill; BCI control performance.

1.     No significant differences were observed between groups for training skill or BCI scores.

2.     The ability of SCI subjects to control robotic arms was not statistically different depending on injury location.

Pfurtscheller et al., 2009




Population: Mean age=41.0±14.5 yr; Gender: males=11, females=4; Time since injury: 86.1 mo; Level of injury: C5 – C12, paraplegia=8, tetraplegia=7; Severity of injury: Not reported.

Intervention: Three types of motor imagery tasks were examined via EEG –based discrimination. Tetraplegic (n=7) and paraplegic (n=8) participants were asked to imagine using their right or left hand. Outcome measures were assessed during and after the tasks.

Outcome Measures: Accuracy (EEG activity).

1.     The average classification accuracy for left versus right hand was 65%.

2.     In five out of eight paraplegic participants, the discrimination accuracy was greater than 70%.

3.     Only one out of seven tetraplegic patients had a discrimination accuracy greater than 70%.

Foldes et al., 2015




Population: Mean age=28 yr; Gender: males=3, females=0; Time since injury: 7 yr; Level of injury: C2=1, C5=2; Severity of injury: AISA A=2, B=1.

Intervention: Patients utilized a BCI for closing and opening a virtual hand to promote hand rehabilitation via therapeutic neuroplasticity. Participants performed 200 trials of hand control movements for approximately 30 min. Outcome measures were assessed after trial completion.

Outcome Measures: BCI performance; Grasp success rate; Grasp or rest sensorimotor rhythms (SMR).

1.     Participants were able to maintain brain-control of closing and opening a virtual hand with a significantly increased success rate of 63% (p<0.001).

2.     Grasp success rates significantly increased for each participant (p<0.001).

3.     Two out of three participants showed significant improvement in SMR (p<0.001), indicating they had learned to change their brain activity with a single session of training.

Pedrocchi et al., 2013




Population: Mean age=52 yr; Gender: males=3, females=0; Time since injury: XX yr; Level of injury: C3 – C7; Severity of injury: AISA Not reported.

Intervention: Participants utilized a Multimodal Neuroprosthesis for daily Upper limb Support (MUNDUS) to perform different tasks related to ADLs, such as reaching and drinking. Outcome measures were assessed by three experts during completion of the task.

Outcome Measures: User intention; Evaluation score (from zero, unsuccessful, to 2, completely functional); Donning time.

1.     The functionality of all modules was successfully demonstrated.

2.     User intention was detected with 100% success.

3.     Averaging all subjects and tasks, the mean evaluation score was 1.6, with a minimum of 1.13.

4.     All users, but one, subjectively perceived the usefulness of the assistance and could easily control the system.

5.     Donning time ranged from 6 to 65 minutes.

Blabe et al., 2015




Population: Age range=15-81; Gender: not reported; Time since injury: <10 yr; Level of injury: C1-C7; Severity of injury: incomplete=90, complete=60.

Intervention: No intervention. A technology survey to determine the likelihood of spinal cord injury patients adopting different technologies, given the burdens currently associated with them.

Outcome Measures: User preference for 8 BMI technologies including EEG, ECoG, intracortical micoelectrode arrays and a commercially available eye tracking system.

1.     Ninety-one percent of respondents with an injury level C1-C4 and 78% of C5-C7 who were <10 yr post injury said they be “likely” to adopt a BMI technology if it could restore some grasp of their hand or restore natural arm movement without sensation. 2.     Control of external devices such as prosthetic (robotic) arms, computer cursors and wheelchairs was of moderately high interest to participants (>60% of C1-C4 respondents).

3.     Participants were less likely to adopt these control capabilities if they were not described as being fast, accurate or natural.

4.     High speed typing and control of a fast prosthetic (robotic) arm were of more interest than restoring less-than-natural native arm movement, via FES.

5.     Surgically implanted wireless technologies were twice as “likely” to be adopted as their wired equivalents.

6.     Thirty-nine percent of patients with C1-4 injury for 10 year or more were likely to adopt wired EEG caps, while 52% of the same population were likely to adopt the wireless intracortical technology.

7.     Forty-eight percent of C1-C4 respondents and 45% of C5-7 respondents with less than 10 yr post injury were likely to adopt the wireless ECoG technology to restore some grasp of the hand, 60% of C1-4 and 46% C5-7 of the same group were likely to adopt wireless intracortical technology if it could restore some grasp of their hand.

8.     Fifty-six percent of C5-7 and 80% of C1-4 respondents were more likely to adopt a technology if it could control a cursor on a computer screen in a completely natural way.

9.     Sixty-four percent of C5-C7 and 72% of C1-4 respondents, would be likely to adopt a technology if it would allow them to type at 40 words per minute with some errors.

Collinger et al., 2013




Population: Mean age=55.2; Gender: male=51, female=6; Time since injury: 10.9 yr; Level of injury: tetraplegia=21, paraplegia=36; Severity of injury: not reported.

Intervention: No intervention. A survey of 57 veterans with SCI to determine priorities in improving quality of life, knowledge of assistive technologies and interest in BCIs.

Outcome Measures: Experience with assistive devices; Functional priorities; BCI technology.

1.     Restoration of bladder, bowel control, walking, and arm and hand function (tetraplegia only) were all high priorities for improving quality of life.

2.     Many of the participants had not used or heard of some currently available technologies designed to improve function or the ability to interact with their environment.

3.     The majority of participants in this study were interested in using a BCI, particularly for controlling functional electrical stimulation to restore lost function.

4.     Independent operation was considered to be the most important design criteria.

5.     Many participants reported that they would consider surgery to implant a BCI even though non-invasiveness was a high-priority design requirement.

Onose et al., 2012




Population: Age range=33.1; Gender: male=8, female=1; Time since injury range: 6-202 mo; Level of injury: C4-C7; Severity of injury: AIS Frankel score one=4, two=3, three=2.

Intervention: Tetraplegic patients assessed the feasibility of a EEG-BCI for reaching/grasping assistance, though a robotic arm and completed a survey.

Outcome Measures: Accuracy; Perception; Side effects.

1.     EEG-BCI performance/calibration-phase classification accuracy averaged 81%; feedback training sessions averaged 70.5% accuracy.

2.     Seven out of nine (77.7%) patients reported having felt control of the cursor and 3 (33.3%) subjects felt they were controlling the robot through their movement imagination.

3.     No significant side effects occurred.

4.     BCI performance was positively correlated with beta EEG spectral power density (p=0.025) and AIS score (p=0.089).


There has been considerable progress in neuroscience and technology, allowing for the development of aids for mobility regeneration. The emergence of neural interface technologies has provided an innovative approach to aid patients with sensorimotor deficits. All of the studies presented in Table 8 demonstrated that the use of BCI technology, although diverse, was feasible. However, the efficacy of BCI technology varied between studies. One randomized controlled trial found that BCI-FES technology not only provided benefit as an assistive device but also improved neurological recovery and muscle strength, possibly through neuroplasticity (Osuagwu et al. 2016). Similarly, Foldes et al. (2015) found that a MEG based BCI improved sensorimotor rhythms to promote neuroplasticity following SCI.

The remainder of articles focused on BCI technology to control external devices. In these studies, it was found that control of a robotic device using BCI technology is feasible and individuals with SCI are interested in using the technology. In a survey that was conducted, 80% of respondents would consider adopting a BCI technology, if it could restore some hand grasp (Blabe et al. 2015). However, it was less likely to be adopted if it was aesthetically unpleasing, unreliable, difficult or embarrassing to use. It should be noted that participant performance on functional tasks was relatively poor. This may be due to the fact that participants needed more time training with the device or that the technology needs to be developed further to provide real benefit for self-assistance. Nonetheless, BCI is a promising rehabilitative device for individuals with SCI.

The importance of BCI applications in the future will depend on their reliability, and technological and functional advantages over conventional technology/rehabilitation. BCI technology has the potential to improve autonomy and independence in basic activities of daily life. For example, simple tasks such as drinking, eating, or moving hair away from the eyes can fundamentally improve quality of life and were identified as the most relevant by a focus group (Collinger et al. 2013). Despite the advantages of this technology, there are some drawbacks including increased donning times, cost and prototype technology that often needs improvement. Future research should focus on determining the long-term effects of BCI use and examine whether this technology could be adapted as a functional rehabilitative device.


There is level 1b evidence (from one randomized controlled trial: Osuagwu et al. 2016) that BCI-FES should be considered as a therapeutic tool rather than solely an assistive device, as combined BCI-FES therapy results in better neurological recovery and muscle strength than FES alone.

There is level 2 evidence (from two prospective controlled trials: Athanasiou et al. 2017; Pfurtscheller et al. 2009) that robotic control of a wireless or EEG controlled BCI is possible in SCI patients, however, multiple training sessions and tailored BCI algorithms are needed to improve performance.

There is level 4 evidence (from one pre-post test: Foldes et al. 2015) that a MEG based BCI may provide realistic, efficient and focused neurofeedback in SCI patients to promote neuroplasticity.

There is level 4 evidence (from one pre-post test: Pedrocchi et al. 2013) that the MUNDUS platform may provide functional assistance in activities of daily living to patients with SCI.

There is level 5 evidence (from two observational studies: Collinger et al. 2013; Blabe et al. 2015) that individuals with SCI are interested in contributing to the design of BCIs and would adopt autonomous BMI systems for control of external devices or the restoration of upper extremity function.

There is level 5 evidence (from one observational study: Onose et al. 2012) that EEG-BCI-mechatronic devices may contribute real but limited potential for self-assistance in individuals with SCI.