Activation Duty Cycle and Regional Muscle Co-Activation Across GMFCS Levels in Ambulatory Children with Spastic Cerebral Palsy: An Exploratory Surface EMG Study.
Huang P, Zhang Z, Tong Y, Sun Y • Sensors (Basel, Switzerland) • 2026
Higher GMFCS level in ambulatory children with spastic cerebral palsy was associated most consistently with longer activation duty cycle, stronger distal and proximal peak-normalized co-activation, and greater inter-muscle spatial entropy, but internal resampling performance did not transfer to an external dataset.
Key Findings
Results
Higher GMFCS level was associated with longer muscle activation duty cycle, stronger distal and proximal peak-normalized co-activation, and greater inter-muscle spatial entropy.
Study population comprised 38 ambulatory children with spastic cerebral palsy at GMFCS levels I–III.
Eight-muscle surface EMG was recorded during self-selected overground walking.
16 peak-normalized envelope and exploratory synergy features were examined.
Directional associations were retained across sensitivity analyses.
Directional concordance was strongest for activation duty cycle and distal co-activation.
Results
The 16-feature pool achieved a quadratic weighted kappa (QWK) of 0.837 ± 0.046 for within-cohort GMFCS discrimination.
Performance was estimated across 50 repeated participant-level nested cross-validation evaluations.
This estimate reflects resampling performance within the 38-child cohort, not independent-population generalizability.
A four-feature selection procedure inside every outer-training fold yielded QWK of 0.849 ± 0.029.
The core-four feature specification was identified post hoc.
Results
Direct transfer of the internally fitted models to an external supplementary dataset was poor.
The supplementary workbook contained 195 unique participant IDs.
Direct transfer of the 16-feature model yielded QWK of 0.210.
Direct transfer of the core-four model yielded QWK of 0.077.
Directional concordance for activation duty cycle and distal co-activation was retained in the external dataset, but model performance was not.
Discussion
The study lacked healthy controls, measured walking speed, and stride length, limiting interpretation of findings as cerebral-palsy-specific biomarkers.
The cross-sectional design identifies descriptors associated with ambulatory GMFCS level rather than CP-specific biomarkers.
Authors explicitly state results do not constitute 'evidence of clinical predictive utility.'
Absence of typically developing controls prevents determination of whether findings are specific to cerebral palsy.
Methods
The study used a design-locked pool of 16 sEMG features to characterize muscle activation patterns across GMFCS levels I–III.
Features included peak-normalized envelope descriptors and exploratory synergy features.
Eight muscles were instrumented per participant.
Participant-level nested cross-validation with correlation reduction and feature selection was applied within outer training folds.
38 ambulatory children with spastic cerebral palsy participated in the study.
What This Means
This research examined muscle activation patterns during walking in 38 children with spastic cerebral palsy who were classified into three levels of mobility (GMFCS levels I–III, where higher numbers indicate greater motor difficulty). Using sensors placed on the skin over eight muscles, the researchers measured how muscles turned on and off during walking and calculated 16 different descriptors of muscle activity. They found that children with higher GMFCS levels — meaning greater walking difficulty — tended to keep their muscles activated for longer portions of each step, showed stronger simultaneous activation of multiple muscle groups both in the legs and around the hips, and had more spatially spread-out patterns of muscle activity.
When the researchers tested how well these 16 muscle-activity features could distinguish between the three GMFCS levels within their own group of 38 children using a rigorous repeated cross-validation method, they obtained strong agreement scores (quadratic weighted kappa of 0.837–0.849). However, when they attempted to apply the same approach to a separate, larger dataset of 195 participants, performance dropped dramatically (kappa of 0.077–0.210), indicating the models did not generalize beyond the original group.
This research suggests that certain sEMG features — particularly how long muscles stay active and how strongly muscle groups co-activate — track with the severity of walking difficulty in children with spastic cerebral palsy. However, because the study lacked healthy comparison children, did not measure walking speed or stride length, and showed poor transfer to an external dataset, the findings should be understood as exploratory associations rather than validated clinical tools. The authors caution that the results identify descriptors linked to GMFCS level but do not establish these as cerebral-palsy-specific biomarkers or demonstrate clinical predictive utility.
Check Your Own Numbers
Upload your bloodwork. We'll cross-reference your results against this study and 4,700 others.
Huang P, Zhang Z, Tong Y, Sun Y. (2026). Activation Duty Cycle and Regional Muscle Co-Activation Across GMFCS Levels in Ambulatory Children with Spastic Cerebral Palsy: An Exploratory Surface EMG Study.. Sensors (Basel, Switzerland). https://doi.org/10.3390/s26175434