Vakgroep Menselijke fysiologie (MFYS)

Recently, technological advancements have led to the use of robotic devices to facilitate gait rehabilitation and assistance in subjects with lower limb impairments and gait disorders. Robot-assisted gait training has a large potential to improve motor function and facilitate walking recovery, while robotic devices for the assistance of gait can augment the walking performance of disabled patients. Yet, to date it is not entirely clear how humans interact with robotic devices for rehabilitation and assistance of the lower limbs and how we can, based on that interaction, maximize the effectiveness of these devices. Until now, emphasis has often been put on the mechanical development and control design of robotic systems and conclusions on its effectiveness are mostly drawn from experiments measuring system performance on one or two subjects. However, in order to optimize the effectiveness of such systems, it is imperative to determine how humans respond to and interact with them. This dissertation covers aspects of the evaluation of robotic assistive devices from a non-engineering viewpoint. More specifically, the purpose of this work was to study the assessment and effects of human-robot interaction (HRI) between a healthy human motor system and robotic devices for rehabilitation and assistance of gait. HRI was investigated in terms of biomechanical and (electro)physiological parameters. All studies were carried out in the framework of the ALTACRO (i.e., Automated Locomotor Training using an Actuated Compliant Robotic Orthosis, http://altacro.vub.ac.be/) and CYBERLEGs project (i.e., The Cybernetic Lower Limb Cognitive Ortho-Prosthesis, http://www.cyberlegs.eu/).

First, a literature review was performed to provide an overview of the state-of-the-art in the assessment of HRI and the effects of robot-aided assistance and rehabilitation of gait on the human body. The review pointed out that the assessment of HRI in the rehabilitation of gait should be aimed at assessing the level of active participation in order to optimize motor learning and improve motor recovery. While evaluating HRI in the assistance of gait, should be focused on assessing walking performance in order to optimize the mobility and independency of the patient. Moreover, HRI can be subdivided into physical (i.e., effects on heart, muscles, brain, …) and cognitive (i.e., effects on emotion, attention, … ) aspects. Further, the review indicated that,  mostly biomechanical (i.e., kinematics, gait parameters, kinetics) and physiological (i.e., muscle activity, heart rate, …) measures have been applied to assess HRI. Recently, also direct measures of cortical engagement, i.e., functional near-infrared spectroscopy (fNIRS) and electroencephalography (EEG), have been introduced. In general, the effects of robotic orthoses and/or prostheses on biomechanical parameters and muscle activity indicate that most of these device are more or less able to provide a gait pattern close to a natural gait pattern. Besides, the effects are largely dependent on the mechanical design and the control system of these devices. With regard to cortical activity, the few studies that have been performed during robot-assisted walking point at the existence of neural correlates of cortical engagement which might be useful for neurofeedback and brain-computer interface (BCI) applications in robot-assisted gait. Nevertheless, several aspects in the assessment and effects of HRI in rehabilitation and assistance of gait remain to be elucidated such as the effects of different levels of assistance on human motor control well as unravelling the contribution of neural correlates of walking and robot-assisted walking to the locomotor control of gait.

In a first study HRI, in terms of muscle activity, kinematics and gait parameters, with a unilateral powered knee exoskeleton (KNEXO) at different levels of compliant assistance was researched. The study showed that healthy subjects can walk with KNEXO in patient-in-charge mode with some slight constraints in kinematics and muscle activity primarily due to inertia of the device. Yet, during robot-in-charge walking the muscular constraints are reversed by adding positive power to the leg swing, compensating in part this inertia. No significant differences in the human response to the interaction with KNEXO in low and high compliant assistance could be pointed. This is in contradiction with our hypothesis that muscle activity would decrease with increasing assistance. It seems that the differences between the parameter settings of low and high compliant control might not be sufficient to observe clear effects in healthy subjects. Moreover, we should take into account that KNEXO is a unilateral, one degree-of-freedom device.

Next, a new technique to assess active participation and performance during robot-assisted gait rehabilitation was explored. In this second study, the feasibility of measuring EEG during a complex movement such as human gait was assessed. The study showed that a characteristic temporal pattern of positive and negative potentials, similar to movement-related cortical potentials (MRCPs), and related to the gait cycle was observed over the cortical leg representation area. Source localization analysis revealed that mainly the primary somatosensory, somatosensory association, primary motor and cingulate cortex were activated during walking. This study demonstrated the feasibility to measure EEG during walking and identified gait-related cortical potentials. The results also indicate a widely distributed cortical network involved in gait control.

Following the feasibility study, EEG was used to assess HRI, in terms of cortical activity, during robot-assisted treadmill walking with the Lokomat-system. The effects of different levels of robotic assistance were explored. The study revealed three active clusters located in the sensorimotor cortex during treadmill walking (TW) and robot-assisted treadmill walking (RATW) in healthy subjects. These clusters demonstrated gait-related modulations in the mu, beta and low gamma bands over the sensorimotor cortex related to specific phases of the gait cycle (i.e., event-related spectral perturbations (ERSP)). Further, mu and beta rhythms were suppressed (i.e., event-related desynchronisations (ERDs)) in the primary sensory cortex during TW compared to RATW with 100% GF, indicating a larger involvement of the sensorimotor area during TW compared to RATW. No significant differences in the spectral power of the mu, beta and low gamma bands could be identified between RATW with different levels of GF. Low levels of guidance force (i.e., GF30, GF60) and thus active participation should be favored when possible during RATW in order to optimize the involvement of sensorimotor areas and stimulate motor learning.

The last study looked at the feasibility of using physiological measures to estimate the cognitive workload during walking. The results showed that breathing frequency and heart rate significantly increased, and heart rate variability significantly decreased with increasing cognitive workload during walking (p<.05). Performance measures (e.g., cadence) only changed under high cognitive workload. As a result, psychophysiological measures are the most sensitive to identify changes in cognitive workload during walking. These parameters reflect the cognitive effort necessary to maintain performance during complex walking and can easily be assessed regardless of the task. This makes them excellent candidates to feed to the control loop of a bio-cooperative prosthesis in order to detect the cognitive workload. This information can then be used to adapt the robotic assistance to the patient’s cognitive abilities.

In conclusion, physiological measures combined with gait parameters can be a valid way to assess the cognitive state of subjects while engaged in walking. This could open doors for the implementation of such measures in bio-cooperative controlled prostheses in order to augment the performance of the wearer. Also, biomechanical and physiological parameters are adequate to assess different aspects of HRI, i.e., human performance and active participation, yet promising results have also been found with a more direct measure of cortical engagement: EEG. Neural correlates of walking detected by means of EEG in this work were, GRCP, ERDs and ERSPs. With regard to the effects of RATW on a healthy human motor system, these studies showed that different levels of compliant assistance in RATW did not alter the level of active participation (i.e., in terms of muscle and brain activity) in healthy subjects. Yet, walking can be distinguished from robot-assisted walking based on neural correlates such as ERDs in healthy subjects. The presence of different neural correlates closely related to the phases of the gait cycle indicates that there is a a cortical contribution to the motor control of human gait.