Background
A cerebral vascular accident is one of the most common causes of walking disabilities, with approximately 60% of the patients suffering from persistent problems in walking [
1]. These impairments are associated with decreased walking speed and stride length [
2], spatial and temporal asymmetry [
2‐
5], and a higher fall risk [
6]. As walking represents a key aspect of independent functioning, regaining safe gait function is one of the main goals in stroke rehabilitation. Robot assisted gait training (RAGT) is a relatively novel approach to gait training, with robotic devices such as the Lokomat (Hocoma AG, Volketswil, Switzerland) now commercially available. The Lokomat is an actuated exoskeleton that guides the limbs through the gait cycle, making it possible to elicit a kinematically normal gait pattern in patients who are incapable of independent stepping [
7]. In order to purposefully use the Lokomat for gait rehabilitation, knowledge is needed on how guided walking in the Lokomat affects the neuromuscular control that underlies hemiparetic gait.
Locomotor training that involves a high number of task specific movement repetitions is generally associated with larger increases in functional gait performance [
8]. In principle, highly intensive gait exercise can be implemented in over-ground or (bodyweight supported) treadmill training, by letting the therapist manually support limb movements. However, such a training approach is strenuous and physically demanding for therapists [
9‐
12], in particular when training patients with a low ambulatory status. By mechanically guiding the limbs through the gait cycle, robotic gait trainers such as the Lokomat provide a less demanding training setting for the therapist, thus allowing patients to make many repetitions of a well-defined, normative gait pattern [
12]. The amount of support (or ‘guidance’) provided by the exoskeleton can be set to the requirements of the patient and the training, but is kept constant throughout the whole gait cycle. Arguably, the experience of successful, normative and symmetric stepping induces task-specific sensory information that may guide locomotor control and inform plastic changes in the central nervous system [
9,
11,
12]. However, offering robotic guidance may also reduce the need for active contribution by the patient, which is an important prerequisite for activity-dependent learning [
13‐
15]. Understanding the potential of the Lokomat for gait training therefore requires knowledge on the extent to which the walker actively contributes to exoskeleton guided gait. To monitor levels of active involvement, studying muscle activity through electromyography (EMG) may be particularly useful. Research on healthy gait suggests that Lokomat guided walking is associated with reduced muscular output compared to regular treadmill walking [
16], and that the amplitude of muscle activity is negatively associated with the amount of guidance that is provided [
17]. Although these findings suggest that guided exoskeleton walking reduces the need for the active control of limb movements, knowledge is lacking on whether the same holds true for patient groups that are targeted for Lokomat training, such as stroke patients..
It is well established that post-stroke hemiparetic gait is altered due to insufficient supraspinal drive, spasticity and peripheral changes in the muscle [
3,
18]. At the neuromuscular level, these aberrations are expressed as abnormal muscular amplitudes and a disrupted temporal ordering of muscle activity, in both the affected and unaffected limb [
3,
19‐
22]. In addition, hemiparetic gait is often characterized by temporal and spatial asymmetry of the stepping pattern [
2‐
5], as patients tend to avoid standing on their affected leg. These abnormalities may require different control strategies in response to the robotic guidance provided by the Lokomat exoskeleton, than previously observed in healthy subjects. However, little is known about the muscle activity that underlies hemiparetic walking in the Lokomat. Only a recent study by Coenen and coworkers [
23] showed that during Lokomat guided walking, muscle activation patterns of stroke patients were more symmetrical, and more similar to healthy treadmill walking. It must be noted that this study involved patients with relatively good walking abilities that were capable of independent overground walking, whereas robot assisted gait training may be specifically indicated for patients with a more severely impaired ambulatory function (see [
7] for a review). Furthermore, no experimental control was exerted over body weight support (BWS) and the levels of exoskeleton guidance, as they were set individually for each patient. Whilst these parameters are known to affect muscle activity [
17,
24‐
26].
The present study wished to elaborate on this work, and assess the difference in muscle activity and temporal control of stepping between Lokomat guided gait and treadmill walking under controlled BWS and Lokomat guidance conditions, in patients with more severe walking problems. More specifically, the aim of this study was threefold. First, we wished to determine how Lokomat guided walking affects muscle activity following stroke and how these effects differ between patients and healthy walkers. To this end, the gait-related muscle activity during Lokomat guided walking and unrestrained treadmill walking were compared between hemiparetic walkers and a group of healthy peers. Second, we aimed to establish how abnormalities in the muscle activity of patients are modulated through Lokomat guided gait. We therefore identified abnormalities during treadmill walking by comparing patients and healthy walkers, allowing us to address how these aberrations are modulated in the Lokomat. Finally, we wanted to determine how temporal step characteristics of patients were modulated during Lokomat guided walking.
Methods
Participants
Ten chronic stroke patients (8 females, 64.4 ± 6.3 years) and ten gender and age-matched healthy controls (7 females, 62.7 ± 4.8 years) volunteered to participate. Patients had a first ever unilateral stroke (infarction or haemorrhage), were at least 3 months post stroke, had unilateral paresis of the leg, and a Functional Ambulation Classification [
27] score of at least 2 (
in our study operationalized as: ‘patient needs continuous or intermittent support of one person to help with balance or coordination’) and at the most 4 (
in our study operationalized as: ‘patient can walk independently in and around the house (<200 m) with help of walking aids, on level ground, but requires help when walking > 200 m, on stairs, slopes and uneven surfaces’). Patients were excluded when they had severely impaired cognitive functions (Mini Mental State Exam [
28] score ≤ 25), severe speech, language or communication disorders, severe visual problems or neglect, or co-morbidities that are known to affect gait or balance performance. Stroke patients were excluded when they were incapable to walk under experimental conditions. None of the healthy subjects suffered from disorders that are known to affect gait performance or muscle activity. None of the participants had previous experience with walking in the Lokomat. For a full overview of the characteristics of the participants, see Table
1.
Table 1
Overview of participant characteristics
Stroke patients |
1 | Female | 57 | 66 | 1.76 | L | Hemorrhage | 5 | 2 | Wheelchair, AFOc
|
2 | Female | 72 | 68 | 1.69 | R | Hemorrhage | 70 | 4 | Cane, AFOc
|
3 | Female | 70 | 70 | 1.63 | L | Infarction | 12 | 3 | Cane |
4 | Female | 72 | 68 | 1.68 | R | Infarction | 216 | 4 | Cane |
5 | Female | 67 | 50 | 1.68 | R | Infarction | 23 | 4 | Cane, one-handed rolling walker, AFOc
|
6 | Female | 55 | 86 | 1.70 | L | Hemorrhage | 53 | 3 | Eifel cane, wheelchair |
7 | Male | 57 | 72 | 1.76 | R | Hemorrhage | 5 | 4 | Cane, adjusted shoes |
8 | Female | 65 | 65 | 1.59 | L | Hemorrhage | 148 | 4 | Rolling walker |
9 | Male | 66 | 91 | 1.83 | L | Infarction | 27 | 4 | Cane, AFOc
|
10 | Female | 63 | 60 | 1.57 | L | Infarction | 4 | 3 | Cane, wheelchair, AFOc
|
Mean (std)
| | 64.4 (6.3) | 69.9 11.8) | 1.69 (0.08) | | | 56,3 (71,5) | | |
Healthy subjects |
Mean (std)
| 7 females | 62.7(4.8) | 77.1(13.0) | 1.74(0.09) | | | | | |
All participants provided their written informed consent. The protocol was in accordance with the Declaration of Helsinki [
29], and approved by the Medical Ethical Committee of the University Medical Center Groningen (METc UMCG, project number: NL46137.042.12), the Netherlands.
Experimental protocol
For this study, the Lokomat Pro version 6.0 (Hocoma AG, Volketswil, Switzerland) was used, which is a bilaterally driven exoskeleton that is combined with a BWS system and a treadmill (for more detailed information see ‘apparatus’ section). The Lokomat was located at the rehabilitation centre ‘Revalidatie Friesland’ in Beetsterzwaag, the Netherlands. Stroke patients visited the rehabilitation centre twice. The first session was not used for testing but only to familiarize patients with the Lokomat, and to evaluate whether they were capable to walk under the experimental walking conditions. The data collection was conducted in the second session (i.e. the test session). Healthy subjects visited the rehabilitation centre only once, for the test session.
During testing the protocol involved two walking conditions, (i) on the Lokomat treadmill, but disengaged from the exoskeleton (henceforth ‘treadmill walking’) and (ii) on the Lokomat treadmill attached to the exoskeleton, that provided 50% guidance (henceforth ‘Lokomat guided walking’). Since muscle activity can be affected by gait speed [
30,
31] and BWS [
24‐
26], these parameters were kept constant throughout the experiment by setting gait speed to 0.56 m/s and providing no BWS for any of the participants. Settings for gait speed and the level of guidance were chosen based on earlier research [
23] and on clinical experience of physiotherapist working with stroke patients in our centre.
To avoid that muscle activity during treadmill walking was confounded by after-effects of Lokomat guided walking, treadmill trails were always conducted before walking in the Lokomat, for all participants. To minimize possible effects of carry-over between walking conditions, and to eliminate possible effects of fatigue, a resting period of at least 5 min was obligatory between the treadmill and Lokomat trial. Patients were allowed extra resting time when needed, until they indicated to be ready for the Lokomat trial. Trial durations were 60 s, and when the Lokomat suddenly stopped, e.g. when unexpected movements triggered the safety mechanism, the trial was repeated until a trial of the required duration was completed. Prior to a trial, participants were allowed practice time, until they indicated to be comfortable with the specific settings. Participants were allowed to rest their hands on the side bars of the Lokomat for stability and wore the BWS harness for safety only, without providing BWS. During Lokomat guided walking ankle movements were stabilised by elastic foot lifters. Patients wore their own (adjusted) footwear and Ankle Foot Orthoses.
Apparatus
The Lokomat Pro
The Lokomat exoskeleton is comprised of two actuated orthoses that are attached to the participant’s limbs by means of cuffs and straps. The geometry (hip width, length of the upper and lower limbs) of the orthoses, and the size and position of leg cuffs were adjusted to the subject’s individual anthropometry, ensuring that walking in the device was as natural and comfortable as possible [
9].
The hip and knee joints of the Lokomat are actuated by linear drives that move the orthoses through the gait cycle, in the sagittal plane [
9,
10,
12], and as such ‘
guide’ the participant’s limbs to move along a predefined path. This predefined pattern is based on joint movements derived from trajectories of healthy walkers [
9,
10], but can be fine-tuned by adjusting the hip- and knee angles to meet walkers functionality. Ankle movements are not actuated, but can be stabilized by elastic foot lifters, to prevent foot drop and concomitant stumbling during the swing phase.
The ‘
guidance’ provided by the Lokomat exoskeleton used in the present study is realized by means of an impedance controller, that allows the level of guidance to be set by the therapist. The level of guidance that is offered determines how much limb movements are permitted to deviate from a predefined pattern. As long as the patient moves along the predefined pattern, the controller does not interfere, but once the limits are exceeded, joint torques are applied to move the limb back towards the desired trajectory [
32]. When guidance is set to its maximum (i.e. 100%), the walker is forced to strictly follow the predefined pattern, but when guidance is set to nil, free limb movements are allowed as exoskeleton torques are applied only to correct for the exoskeleton inertia [
10,
12,
17].
In the present study, the level of guidance during the ‘Lokomat guided walking’ condition was set to 50%, which allows small deviations and requires more active involvement of the walker compared to fully guided walking.
Electromyography and detection of gait events
To assess muscle activity, surface EMG was used to measure activity of the Gluteus Medius (GM), Vastus Lateralis (VL), Biceps Femoris (BF), Medial Gastrocnemius (MG) and Tibialis Anterior (TA). Since stroke patients are known to display abnormal neuromuscular control in both the affected and unaffected limb [
3,
19‐
22], EMG was measured bilaterally in patients. However, in healthy participants only EMG of the dominant leg was measured, as Ôupuu and Winter [
33] showed that in a group of healthy walkers EMG does not differ between the dominant and non-dominant leg. Signals were recorded using self-adhesive, disposable Ag/AgCl electrodes (Kendall/Tyco ARBO; Warren, MI, USA) with a 10 mm diameter and a minimum electrode distance of 25 mm. Sensor placement conformed to the SENIAM guidelines [
34]. To improve skin conduction, the electrode sites were prepared by removing body hair, and by abrading and cleaning the skin with alcohol.
Custom-made insoles equipped with 4 pressure sensors (FSR402, diameter 18 mm, loading 10 – 1000 g, one under the heel and three under the forefoot) were used to detect initial foot contact and swing onset for both legs. Pressure sensor and EMG signals were simultaneously sampled at 2048 Hz and fed to a Porti7 portable recording system (Twente Medical Systems, Enschede, The Netherlands). The unit (common mode rejection of >90 dB, a 2μVpp noise level and an input impedance >1 GV) pre-amplified and A/D converted (22 bits) the signals before storage on a computer for offline analysis.
Data analysis
Signal analysis
Offline analysis of pressure sensor and EMG data was done using costum-made software routines in Matlab (version 2015b; The Mathworks Inc., Natick, MA). In line with SENIAM recommendations [
35], EMG data were firstly high-pass filtered using a 4
th order Butterworth with a cut-off frequency of 10 Hz, to reduce movement artefacts. Subsequently, the data were full wave rectified and low-pass filtered (10 Hz 4
th order Butterworth). Pressure sensor data were used to distinguish the first double support (DS1), the single support (SS), the second double support (DS2) and the swing (SW) phase, for both legs. The summed (rectified and low-pass filtered) EMG data was calculated for each of these sub-phases and subsequently averaged over all strides, for each participant and each condition, for further statistical processing. For visual presentation of the data only, the filtered EMG data of each individual step were time-normalized with respect to gait cycle time (i.e. 0 – 100%, heelstrike to heelstrike), and subsequently averaged over strides. To assess the temporal structure of the hemiparetic gait pattern, relative durations (expressed as a percentage of the total gait cycle time) of the DS1 and SS phase were calculated and averaged over strides for each of the limbs, of stroke patients only. Similar to earlier patient-related EMG studies [
19,
23,
36], no amplitude normalization was performed on the EMG data. Normalization of EMG amplitude reduces the inter-individual variation in EMG amplitude by dividing the measured amplitude in microvolts by e.g. the maximum amplitude measured over all conditions, so that EMG values are expressed as a percentage of the maximal amplitude. A potential limitation of this procedure is that, if the signal to noise ratio is low (e.g. when the overall amplitude of the signal is low, which is not uncommon following stroke [
37,
38]), background noise will have a disproportional contribution to the amplitude normalized signal, which may result in unreliable group averages. Therefore, it was chosen not to normalize the EMG amplitude.
Statistical analysis
To compare levels of muscle activity between treadmill walking and Lokomat guided walking (within subjects factor ‘Condition’), and to determine whether the effects of walking condition differed between stroke patients and healthy walkers (between subjects factor ‘Group’), a repeated measures ANOVA was conducted separately for each of 4 sub-phases (DS1, SS, DS2, and SW).
To elucidate how aberrant reactions of patients to Lokomat guided walking were related to abnormalities as displayed during unrestrained treadmill walking, a supplementary analysis was conducted. In case of a significant Condition by Group interaction (implying that differences between Lokomat guided walking and treadmill walking were dissimilar for patients and healthy walkers), independent sample t-tests were conducted to compare the groups during treadmill walking. This way, abnormalities in the EMG of patients were determined for unrestrained treadmill walking, allowing us to specifically address how these abnormalities (i.e. abnormally high or abnormally low activity) were modulated during Lokomat guided walking.
To determine whether temporal step asymmetries in patients were modulated during Lokomat guided walking, a Repeated Measurements ANOVA was conducted on the duration of the DS1 and SS phase, to compare the effects of the within subject factors ‘Limb’ (affected vs unaffected leg) and ‘Condition’ (treadmill walking vs. Lokomat guided walking). We specifically focused in the Limb by Condition interactions, indicating that temporal asymmetry in patients was altered through Lokomat guided walking.
Statistical analyses were performed with SPSS version 20 for Windows (SPSS, Chicago, IL,USA). All test results were evaluated with an alpha of 5%, and the Benjamini-Hochberg correction was used to correct for multiple testing [
39]. All analyses were done separately for each gait phase.
Acknowledgements
The authors firstly wish to thank the participants, and Steven Floor, Marije Huisinga and Sylvana Weiland for their help during data collection. Secondly, the authors wish to express their gratitude to prof. dr. Chris Visscher, Stefan Luis and Emyl Smid for their support and advice.