Page 152 - Handbook of Biomechatronics

Biomechatronic Applications of Brain-Computer Interfaces 149

performance. Furthermore, the competition emphasized the importance of
effective BCI training for the user—the teams all had very different
participant-training strategies, and the winning team stated that their effec-
tive BCI training regimen (which included mock audiences and loud noises)
likely had a major effect on their success (Perdikis et al., 2017).

2.3 Control of Artificial Limbs

Artificial limbs that can be controlled using only brain signals are a staple of
science fiction and would be extremely useful for amputees. State-of-the-art
powered limb prostheses are generally controlled by the EMG of residual
muscles, but often include unintuitive and complicated control schemes that
require significant user training, which limits user acceptance (Farina et al.,
2014). BCI-controlled prostheses could be significantly more unintuitive, as
they could directly interpret desired commands from the motor cortex,
making the user feel as if they are controlling their own limb. A step in this
direction, but without BCIs, was taken by the surgical technique of targeted
muscle reinnervation: motor nerves that previously led from the brain to the
missing limb are surgically reattached to a different muscle, controlling that
muscle’s behavior, and the EMG of that muscle is then used to control the
prosthesis (Cheesborough et al., 2015). However, BCIs could streamline the
process further by directly connecting the brain to the prosthetic limb.
Unfortunately, noninvasive BCI methods are too inaccurate, unin-
tuitive, and/or nonportable for control of artificial limbs. SSVEPs and
P300 responses, which rely on an additional screen to provide visual stimuli,
cannot be used with a prosthetic limb due to mobility issues, though they
could be used with a fixed artificial limb such as a robotic arm that is attached
to a dinner table and assists with self-feeding. For example, Ortner et al.
(2011) developed an assistive orthosis that moved a paralyzed user’s arm
via SSVEP control. The system was tested with participants with tetraplegia
and achieved reasonable performance rates, though participants complained
about the flickering lights required to evoke SSVEP responses.
Both motor and mental imagery could, in principle, be used with pros-
thetic arms and have actually been used to control the behavior of a stationary
robotic arm (Hortal et al., 2015). BCI users were successfully able to pick up
boxes and move them to a different location using the arm, but the classifi-
cation accuracy was relatively low—significantly worse than state-of-the-art
EMG-based prosthesis control. In a related study, motor imagery was com-
bined with SSVEPs for robotic arm control: imagery was used to open and