Page 168 - Handbook of Biomechatronics

Biomechatronic Applications of Brain-Computer Interfaces 165

participants (Sp€uler et al., 2012), others have found essentially no difference
(Leeb et al., 2015), and it is not clear how different pathologies affect per-
formance in different tasks.
Determining the effect of individual characteristics on BCI performance
in different tasks is admittedly a daunting task, as it would require multiple
studies (due to the need for different tasks) and many participants per study
(since the effects of many characteristics would likely be small). The most
efficient way to obtain this information may be through a focused review
paper that would combine information from many studies to obtain a bigger
picture of these effects.

3.3 Training Regimens and User-BCI Coadaptation

BCI performance tends to improve as users train with the system (Lotte
et al., 2013; Neuper and Pfurtscheller, 2010). However, this is not a simple
dose-response relationship: it is a complex process of the user and machine
learning to adapt to each other’s idiosyncrasies. Thus, while interacting with
a BCI, users will develop their own strategies to compensate for BCI imper-
fections. For example, in our recent interviews of participants at the
Cybathlon 2016 BCI race (Novak et al., 2018), we noted that participants
were aware of the delay in detection of motor imagery (up to a second
between imagining the motion and the BCI sending a command in response
to the detected imagery), and compensated for it by imagining the motion
before the command actually needed to be sent. While this led to premature
command triggers and consequent penalties in some participants, it was able
to improve BCI performance for other participants who were able to master
the required prediction process. However, this prediction was not learned
instantly: it was part of the BCI training process that, in some participants,
involved over a hundred practice races.
As another Cybathlon example, all participants were aware that the
actual Cybathlon BCI competition would involve racing in a highly stressful
environment with thousands of noisy spectators and that it would not be
possible to tailor the BCI to that environment through laboratory training.
To make the training more relevant, some participating teams simulated the
competition environment in their laboratory using smaller teams of noisy
spectators (Novak et al., 2018). Furthermore, after the event, some teams
complained about unexpected factors that may have negatively affected their
performance, such as increased electromagnetic noise in the environment
due to thousands of cellphones and other devices. These examples illustrate