Abstract:
Operation of assistive exoskeletons based on voluntary control of sensorimotor rhythms (SMR, 8-12 Hz) enables intuitive control of finger or arm movements in severe paralysis after chronic stroke or cervical spinal cord injury (SCI). To improve reliability of such systems outside the laboratory, in particular when brain activity is recorded non-invasively with scalp electroencephalography (EEG), a hybrid EEG/electrooculography (EOG) brain/neural-machine interface (B/NMI) was recently introduced. Besides providing assistance, recent studies indicate that repeated use of such systems can trigger neural recovery.
However, important prerequisites have to achieved before broader use in clinical settings or everyday life environments is feasible. Current B/NMI systems predominantly restore hand function, but do not allow simultaneous control of more proximal joints for whole-arm motor coordination as required for most stroke survivors suffering from paralysis in the entire upper limb. Besides paralysis, cognitive impairments including post-stroke fatigue due to the brain lesion reduce the capacity to maintain effortful B/NMI control over a longer period of time. This impedes the applicability in daily life assistance and might even limits the efficacy of neurorehabilitation training. In contrast to stroke survivors, tetraplegics due to cervical SCI lack motor function in both hands. Given that most activities of daily living (ADL) involve bimanual manipulation, e.g., to open the lid of a bottle, bilateral exoskeleton control is required but was not shown yet in tetraplegics.
To further enhance B/NMI systems, we first investigated whether B/NMI whole-arm exoskeleton control in hemiplegia after chronic stroke is feasible and safe. In contrast to simple grasping, control of more complex tasks involving the entire upper limb was not feasible with established B/NMIs because high- dimensionality of such multiple joint systems exceeds the bandwidth of these interfaces. Thus, we blended B/NMI control with vision-guidance to receive a semiautonomous whole-arm exoskeleton control. Such setup allowed to divide ADL tasks into a sequence of EEG/EOG-triggered sub-tasks reducing complexity for the user. While, for instance, a drinking task was resolved into EOG-induced reaching, lifting and placing back the cup, grasping and releasing movements were based on intuitive SMR control. Feasibility of such shared vision-guided B/NMI control was assumed when executions were initialized within 3 s (fluent control) and a minimum of 75 % of subtasks were executed within that time (reliable control). We showed feasibility in healthy subjects as well as stroke survivors without report of any side effects documenting safe use.
Similarly, feasibility and safety of bilateral B/NMI control after cervical SCI was evaluated. To enable bilateral B/NMI control, established EEG-based grasping and EOG-based releasing or stop commands were complemented with a novel EOG command allowing to switch laterality by performing prolonged horizontal eye movements (>1 s) to the left or to the right. Study results with healthy subjects and tetraplegics document fluent initialization of grasping motions below 3 s as well as safe use as unintended grasping could be stopped before a full motion was conducted. Superiority of novel bilateral control was documented by a higher accuracy of up to 22 % in tetraplegics compared to a bilateral control without prolonged EOG command.
Lastly, as reliable B/NMI control is cognitively demanding, e.g., by imagining or attempting the desired movements, we investigated whether heart rate variability (HRV) can be used as biomarker to predict declining control performance, which is often reported in stroke survivors due to their cognitive impairments. Referring to the close brain-heart connection, we showed in healthy subjects that a decline in HRV is specific as well as predictive to a decline in B/NMI control performance within a single training session. The predictive link was revealed by a Granger-causality analysis.
In conclusion, we could demonstrate important enhancements in B/NMI control paradigms including complex whole-arm exoskeleton control as well as individual performance monitoring within a training session based on HRV. Both achievements contribute to broaden the use as a standard therapy in stroke neurorehabilitation. Especially the predictive characteristic of HRV paves the way for adaptive B/NMI control paradigms to account for individual differences among impaired stroke survivors. Moreover, we also showed feasibility and safety of a novel implementation for bilateral B/NMI control, which is necessary for reliable operation of two hand-exoskeletons for bimanual ADLs after SCI.
brain-computer interface
Print-on-Demand