Page 157 - Handbook of Biomechatronics

154 Domen Novak

2.6 BCI-Triggered Motor Rehabilitation
In motor rehabilitation after stroke, spinal cord injury, traumatic brain
injury, or other diseases, patients must perform repetitive, intensive limb
exercise to regain their motor functions. Such therapy is increasingly fre-
quently provided by rehabilitation robots that hold the patient’s limb and
assist in making the desired motion (Lo et al., 2010; Klamroth-Marganska
et al., 2014). However, even if the robot provides assistance, the motion
should be initiated by the patient, as this allows a tighter coupling between
the motor plan in the cortex and its execution through the robot, thus better
promoting brain plasticity after the injury (Muralidharan et al., 2011). In
patients who still have some residual motion ability, this motion initiation
can be detected by a change in limb position (i.e., the robot does not start
assisting until the patient has moved their limb at least a little) or by measur-
ing limb EMG, which appears before the actual change in limb position and
thus allows a faster robot response (Dipietro et al., 2005). However, these
approaches are not feasible for patients who have no residual motion ability.
In such severely paralyzed patients, we can instead use a BCI to detect
desired motion initiation and have the rehabilitation robot react to it.
BCIs for detection of motion initiation are based on motor imagery: the
patient imagines moving the limb that is undergoing rehabilitation, and this
imagery is decoded with the same approaches used for, for example, control
of mobile robots, then used to trigger a rehabilitation robot that helps carry
out the motion. An early clinical demonstration of this approach was per-
formed by Ramos-Murguialday et al. (2013), who divided patients with
severe upper limb impairment (no ability to move on their own) into
two groups that both participated in 18days of training. In the experimen-
tal group, patients imagined moving their limb, and a hand-and-arm orthosis
then moved the limb in response to detected motor imagery. In the control
group, the hand-and-arm orthosis performed the same amount of limb
motion in a session, but the motions occurred at random times that had
no relation to patient intentions. The experimental group exhibited signif-
icantly higher increases in standard scores of functional arm ability, indicat-
ing that providing proprioceptive feedback that is contingent upon control
of sensorimotor brain activity may improve the beneficial effects of
physiotherapy.
Following the Ramos-Murguialday study, several research groups have
performed clinical evaluations of BCI-triggered motor rehabilitation,
though with mixed results. For example, Ang et al. evaluated robot-aided
rehabilitation with and without a BCI using two robotic systems: the