The aim of this study was to compare gait phase characteristics and muscle activation patterns during different rehabilitative walking conditions (DGO walking and treadmill walking) in groups of healthy children and children with neuro-orthopedic gait disorders.
Our main findings were: (i) differences in kinematics (duration of stance and swing phase) between young patients and healthy participants were found; (ii) Muscle activation patterns of young patients with neuro-orthopedic movement disorders who walked in DGO conditions resembled reference curves of healthy children walking over-ground in general well; (iii) In addition, muscle activation patterns were quite similar between the healthy participants and the young patients during walking in the DGO; (iv) Encouragement of the therapist during DGO results in higher sEMG amplitudes compared to DGO without motivation in patients.
Stance phase duration can be influenced during robot-assisted gait training
For persons unfamiliar with RAGT, it might come as a surprise that step duration can actually be influenced by patients and healthy participants when walking in a DGO. Despite that the device used in this study (Lokomat®), is position controlled (i.e. there is a fixed relationship between position of the legs and time during the gait cycle), differences in relative stance and swing duration were observed between healthy children and those with neuro-orthopedic diagnoses. Compared to the healthy participants, the patients had a prolonged stance phase duration. Although the relative duration of the stance phase is larger at slower speeds , it is unlikely that speed was the underlying factor here, as the walking speed in the DGO condition was similar for both groups (healthy participants: 1.70 ± 0.05 km/h; patients: 1.70 ± 0.20 km/h). During treadmill walking, differences in walking speed might have contributed to the difference in stance duration, as the patients walked slower (1.3 ± 0.4 km/h) compared to the healthy participants (1.7 ± 0.05 km/h).
Nevertheless, during DGO, the relative duration of the stance phase of our patients resembled already reported percentages for healthy children walking over-ground well. Granata et al.  showed in 11 healthy children (mean age 6.5 years) who walked over-ground that the stance duration amounted to 58.7% ± 2.6% (mean ± SD) of the gait cycle, despite a higher walking speed (4.28 ± 1.12 km/h). In addition, Chang et al.  showed in 26 healthy children (mean age 14.7 years) that the stance duration over-ground amounted to 60% ± 2%. During unassisted treadmill walking, patients and healthy participants spent considerably more time in the stance phase compared to these values.
Comparing muscle activation patterns during the DGO conditions and treadmill walking with reference patterns
For clinical practice it appears important to promote exploration of movement strategies, therefore to train on restorative rather than compensatory mechanisms . Furthermore, to induce functionally relevant plastic changes in the brain, training should be task-specific, because brain plasticity in human locomotor networks seems to be task-dependent as well . Indeed, both treadmill walking and RAGT can be considered forms of task-dependent training.
Interestingly, in our study several results indicated that RAGT could induce a physiological walking pattern (aiming at restoration) in children with neuro-orthopedic movement disorders. When comparing our muscle activation patterns to published data from Chang et al. , we noticed that the DGO walking conditions correlated better to walking over-ground than treadmill walking, both in healthy children and patients. Particularly, the DGO condition without encouragement led to the most physiological muscle activation patterns. In patients, however, additional motivation of the therapist did not deteriorate the muscle activation pattern, and as muscle activation amplitudes were higher with encouragement, this condition might actually be favored when training young patients with neuro-orthopedic movement disorders. While the sEMG amplitudes during additional encouragement were higher in healthy children, the sEMG patterns correlated less well with the normal DGO walking condition (mainly in thigh muscles during swing). We assume that this might have been caused by the ability of the healthy children to generate excessive muscle strength, but at inappropriate time-points during the gait cycle against the robot. This might not have occurred in the patients, as they still required proper guidance of the movement by the DGO. Consequently, it is important to underline that the motivational instructions should not only be hortative but also specific to gait events. The smallest correlation with muscle activation patterns from healthy children walking over-ground was found for the VM muscle activity during swing. VM activity was prolonged in patients. We consider this less important, as the main function of VM is to extend and stabilize the knee during stance.
When we compared patients walking on the treadmill with the healthy over-ground reference muscle activation patterns, these patterns were substantially different, especially during the stance phase. We assume that when therapists would have manually assisted the walking pattern of patients during treadmill walking, kinematics and EMG patterns might have been more physiological. Domingo et al.  could observe this for adult patients with incomplete spinal cord injury, but only for the VM muscles and especially at higher speeds, whereas they mentioned there that it would be difficult for the patients anyway to walk at fast speeds without assistance. Another reason might be the increased metabolic costs for patients when walking without passive guidance . Nevertheless, we also observed poor correlations for the healthy participants during treadmill walking and the healthy reference values. We are not sure what might have caused these substantial differences, as visually walking on the treadmill appeared normal. However, even healthy children are known to walk with high variability in the pattern of muscular activation ; within session sEMG variability in children aged 6–8 years was twice as high as reported in adults . Chang et al.  found in children that about 13% of the sEMG curves were not functionally interpretable as physiological gait patterns. Both muscle activation patterns and stride to stride variability showed substantial variability  and stride to stride variability is higher in patients with neurological impairments . Furthermore, walking speed was relatively slow. It is unlikely that slow speed itself might have influenced muscle activation patterns, as these remain relative stable, while the amplitudes change substantially . Only for very slow walking speeds (0.06 m/s; 0.2 km/h), additional bursts can be observed . However, the slow walking speed might have increased balance requirements as all children had to keep (slight) hand contact with the parallel bars next to the treadmill. Finally, all children were still wearing the harness during treadmill walking (without bodyweight support). Both factors might have influenced the walking pattern and therewith muscular activation patterns.
Comparing muscle activation patterns during different conditions
In TA, the typical onset of activation starts before toe-off continuing with full swing phase up to heel strike and loading (about 55-15% gait cycle). We observed TA activity up to approximately 40% of the stance phase. This has been reported previously (e.g. ) and was explained by the activity of TA as a foot inverter muscle to control balance during single support and contra lateral limb swing . Abnormal silence of the TA muscle in terminal swing was reported in patients with lengthened Achilles tendon after clubfoot surgery as well as prolonged GM muscles . This effect was also visible in our study, while most physiological TA activity in late swing could be determined in both groups during DGO walking with motivation.
Nevertheless, TA activity in DGO and treadmill walking appeared more silent in the loading response and the terminal swing compared to normal. We assume that the presence of foot-lifers during DGO enabled good foot clearance during the swing phase and might have facilitated eccentric muscle control during heel strike. Except for the stance phase in patients, we could not observe a significant lower TA activity during DGO compared to treadmill walking. Lower TA activity levels in the DGO in adults have been reported previously  and were explained by the use of foot-lifters. We could not observe this as that much, potentially because we tried to adjust the tension of the foot-lifers to the needs of the child; enough for good foot clearance during swing, but not too strong to make the ankle joint stiff.
Normal activation of GM muscle starts at mid-swing, develops to a maximum at terminal stance and pre-swing (approximately 15-50% of the gait cycle) and is silent during swing. Our results show an early onset of GM activity during the end of swing, as well as a prolonged activity in stance. This is also known as the plantar flexion-knee extension couple to control the second rocker and an upright position. We found this especially in healthy children, mainly during treadmill walking and DGO with motivation condition. Especially in patients, GM amplitudes were small. This could be a consequence of the 30% body-weight-support during DGO walking, which might have reduced the anti-gravitational activity of already weakened GM muscles.
Best GM muscle activity pattern in our study could be found in both groups during DGO walking and in patients with neuro-orthopedic disorders also during DGO with motivation.
Normally, VM is active from mid-swing to mid-stance (75-30% gait cycle). However, during treadmill condition in our study, especially the children with neuro-orthopedic impairments showed activation in terminal stance, which might indicate a co-contraction for stabilization the knee joint before entering in pre-swing. Similar results were observed in the study of . It is noticeable that the VM activity during treadmill walking was higher in children with neuro-orthopedic disorders than in the healthy children. This could be a consequence of the suboptimal gait pattern requiring higher muscle activity. This finding has also previously been reported by Lauer et al. , who compared rectus femoris and medial hamstrings activity between younger children and older ones, as well as between typical developed children and children with cerebral palsy. In typically developing children, older children had elevated muscle activity compared to the younger ones, while children with cerebral palsy showed much higher activity levels in the younger ones, especially in rectus femoris. Another finding was that the VM activity was very variable in the patients. Nevertheless, while the importance of quadriceps muscles is known in gait rehabilitation, it is nice to see that VM activity could be visually observed as most physiological in patients with neuro-orthopedic disorders during DGO as well as DGO with motivation conditions.
The most frequent activation modality of BF starts during mid-swing and continues up to mid-stance (85-10% gait cycle). In our results, BF was exceptionally silent in the DGO and treadmill condition, but highly activated during DGO with additional therapist motivation, mainly in late loading and mid-stance as well as in terminal swing. This could be explained by the excessive backward push of the participants’ leg after heel strike and the resistance of the DGO to this movement, which was also observable in the study of Hidler and Wall .
Encouragement increases muscle activation without affecting the pattern in patients
We expected that therapeutic motivation could increase muscle activation without changing the muscle activation patterns in their shape. This could be confirmed in all four muscles during the whole gait cycle in children with neuro-orthopedic gait disorders. Similarly, healthy controls could increase muscle activation during RAGT with additional encouragement, except for GM and VM during stance. In contrast to the patients, however, the muscle activation patterns changed considerably for GM, VM and BF (the latter only during the swing phase).
This study leaves some space for improvement in either design or data acquired. First, the number of children is relatively small and the group is relative inhomogeneous. Nevertheless, these children represent the patient population that a pediatric rehabilitation center has.
Second, although at least 2 minutes were given to familiarize the participants with the treadmill or 5 minutes for the robotic device respectively, this may have been not enough to ensure habituation and could have influenced the gait pattern.
Third, due to practical limitations in the test protocol, treadmill walking was always recorded at the end of the procedure, which might have caused some fatigue. On the one hand, this might explain why we found hardly any differences between muscle activity amplitudes between the DGO and the treadmill condition, despite that during DGO, 30% bodyweight support was provided. On the other hand, a break was provided before treadmill walking started. In addition, patients spent less time walking in the DGO compared to a regular clinical training session. Furthermore, even for the healthy participants, we found hardly any differences between the DGO and the treadmill condition. Therefore, it is unlikely that these experienced fatigue, as they walked at considerable slower speeds compared to normal. Finally, we did not investigate muscle activation patterns and kinematics during treadmill walking with therapeutic encouragement. This could also be considered a limitation of this study.
Fourth, especially during treadmill walking, it was sometimes difficult to trigger “heel strike” and “toe off”, as this was performed manually through video synchronization. Especially in these patients, the normal heel-toe gait pattern is often variable or absent, which forced us to use video-synchronization rather than foot-switches.
Fifth, while sEMG data were gathered with 2000 Hz, video recordings were made with 50 Hz only. The reduced sampling rate of the video recordings might have influenced the accuracy of determining stance and swing phase, however, due to the low walking speed of the children this is not a critical issue. Moreover, results will not be affected differently between patients and healthy participants, because both walked at equal speeds during DGO conditions, but it might have influenced the results obtained during treadmill walking.