Background: Walking problems in children with cerebral palsy (CP) can in part be explained by limited selective motor control. Muscle synergy analysis is increasingly used to quantify altered neuromuscular control during walking. The early brain injury in children with CP may lead to a different development of muscle synergies compared to typically developing (TD) children, which might characterize the abnormal walking patterns. Objective: The overarching aim of this review is to give an overview of the existing studies investigating muscle synergies during walking in children with CP compared to TD children. The main focus is on how muscle synergies differ between children with CP and TD children, and we examine the potential of muscle synergies as a measure to quantify and predict treatment outcomes. Methods: Bibliographic databases were searched by two independent reviewers up to 22 April 2019. Studies were included if the focus was on muscle synergies of the lower limbs during walking, obtained by a matrix factorization algorithm, in children with CP. Results: The majority (n = 12) of the 16 included studies found that children with CP recruited fewer muscle synergies during walking compared to TD children, and several studies (n = 8) showed that either the spatial or temporal structure of the muscle synergies differed between children with CP and TD children. Variability within and between subjects was larger in children with CP than in TD children, especially in more involved children. Muscle synergy characteristics before treatments to improve walking function could predict treatment outcomes (n = 3). Only minimal changes in synergies were found after treatment. Conclusions: The findings in this systematic review support the idea that children with CP use a simpler motor control strategy compared to TD children. The use of muscle synergy analysis as a clinical tool to quantify altered neuromuscular control and predict clinical outcomes seems promising. Further investigation on this topic is necessary, and the use of muscle synergies as a target for development of novel therapies in children with CP could be explored.

Purpose We sought to identify the developing maturity of walking and running in young children. We assessed gait patterns for the presence of flight and double support phases complemented by mechanical energetics. The corresponding classification outcomes were contrasted via a shotgun approach involving several potentially informative gait characteristics. A subsequent clustering turned out very effective to classify the degree of gait maturity. Methods Participants (22 typically developing children aged 2–9 years and 7 young, healthy adults) walked/ran on a treadmill at comfortable speeds. We determined double support and flight phases and the relationship between potential and kinetic energy oscillations of the center-of-mass. Based on the literature, we further incorporated a total of 93 gait characteristics (including the above-mentioned ones) and employed multivariate statistics comprising principal component analysis for data compression and hierarchical clustering for classification. Results While the ability to run including a flight phase increased with age, the flight phase did not reach 20% of the gait cycle. It seems that children use a walk-run-strategy when learning to run. Yet, the correlation strength between potential and kinetic energies saturated and so did the amount of recovered mechanical energy. Clustering the set of gait characteristics allowed for classifying gait in more detail. This defines a metric for maturity in terms of deviations from adult gait, which disagrees with chronological age. Conclusions The degree of gait maturity estimated statistically using various gait characteristics does not always relate directly to the chronological age of the child.

Muscle synergies reflect the presence of a common neural input to multiple muscles. Steering small sets of synergies is commonly believed to simplify the control of complex motor tasks like walking and running. When these locomotor patterns emerge, it is likely that synergies emerge as well. We hence hypothesized that in children learning to run the number of accompanying synergies increases and that some of the synergies’ activities display a temporal shift related to a reduced stance phase as observed in adults. We investigated the development of locomotion in 23 children aged 2–9 years of age and compared them with seven young adults. Muscle activity of 15 bilateral leg, trunk, and arm muscles, ground reaction forces, and kinematics were recorded during comfortable treadmill walking and running, followed by a muscle synergy analysis. We found that toddlers (2–3.5 years) and preschoolers (3.5–6.5 years) utilize a “walk-run strategy” when learning to run: they managed the fastest speeds on the treadmill by combining double support (DS) and flight phases (FPs). In particular the activity duration of the medial gastrocnemius muscle was weakly correlated with age. The number of synergies across groups and conditions needed to cover sufficient data variation ranged between four and eight. The number of synergies tended to be smaller in toddlers than it did in preschoolers and school-age children but the adults had the lowest number for both conditions. Against our expectations, the age groups did not differ significantly in the timing or duration of synergies. We believe that the increase in the number of muscle synergies in older children relates to motor learning and exploration. The ability to run with a FP is clearly associated with an increase in the number of muscle synergies.

The first years of life might be critical for encouraging independent walking in children with cerebral palsy (CP). We sought to identify mechanisms that may underlie the impaired development of walking in three young children with early brain lesions, at high risk of CP, via comprehensive instrumented longitudinal assessments of locomotor patterns and muscle activation during walking. We followed three children (P1–P3) with early brain lesions, at high risk of CP, during five consecutive gait analysis sessions covering a period of 1 to 2 years, starting before the onset of independent walking, and including the session during the first independent steps. In the course of the study, P1 did not develop CP, P2 was diagnosed with unilateral and P3 with bilateral CP. We monitored the early development of locomotor patterns over time via spatiotemporal gait parameters, intersegmental coordination (estimated via principal component analysis), electromyography activity, and muscle synergies (determined from 11 bilateral muscles via nonnegative matrix factorization). P1 and P2 started to walk independently at the corrected age of 14 and 22 months, respectively. In both of them, spatiotemporal gait parameters, intersegmental coordination, muscle activation patterns, and muscle synergy structure changed from supported to independent walking, although to a lesser extent when unilateral CP was diagnosed (P2), especially for the most affected leg. The child with bilateral CP (P3) did not develop independent walking, and all the parameters did not change over time. Our exploratory longitudinal study revealed differences in maturation of locomotor patterns between children with divergent developmental trajectories. We succeeded in identifying mechanisms that may underlie impaired walking development in very young children at high risk of CP. When verified in larger sample sizes, our approach may be considered a means to improve prognosis and to pinpoint possible targets for early intervention.

Accelerometers are low-cost measurement devices that can readily be used outside the lab. However, determining isolated gait events from accelerometer signals, especially foot-off events during running, is an open problem. We outline a two-step approach where machine learning serves to predict vertical ground reaction forces from accelerometer signals, followed by force-based event detection. We collected shank accelerometer signals and ground reaction forces from 21 adults during comfortable walking and running on an instrumented treadmill. We trained one common reservoir computer using segmented data using both walking and running data. Despite being trained on just a small number of strides, this reservoir computer predicted vertical ground reaction forces in continuous gait with high quality. The subsequent foot contact and foot off event detection proved highly accurate when compared to the gold standard based on co-registered ground reaction forces. Our proof-of-concept illustrates the capacity of combining accelerometry with machine learning for detecting isolated gait events irrespective of mode of locomotion.