Machine learning algorithms for activity recognition in ambulant children and adolescents with cerebral palsy.
Matthew Ahmadi, Margaret O'Neil, Maria Fragala-Pinkham and 2 others
PMID 30442154WHAT IT FOUND
Machine learning models accurately identified brisk walking and sedentary time in children with cerebral palsy using wrist or hip sensors.
Combining both locations improved accuracy. Comfortable walking was often misclassified as brisk walking, so current models cannot reliably distinguish walking speeds.
Key findings
01Classifiers achieved excellent accuracy for sedentary activities (94.1–97.9%) and good to excellent accuracy for brisk walking (71.5–86.0%), but modest accuracy for comfortable walking (47.6–70.4%).
02Random Forest and Support Vector Machine algorithms showed significantly better classification accuracy than Binary Decision Trees across all sensor placements.
03Classifiers trained on combined hip and wrist data consistently provided higher recognition accuracy than those using a single sensor location.
STILL TO COME
How it was doneWhat they foundWhat it means for PTsWhat it means for OTs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The models were trained and tested only on controlled, short-duration activity trials, not in real-world free-living conditions. Only two participants had GMFCS level III mobility, so the results may not apply to children with more severe motor impairments. Comfortable walking was frequently misclassified as brisk walking, partly because the brisk walking speeds of children with severe impairments overlapped with the comfortable walking speeds of children with less severe impairments. The study did not predict energy expenditure, only activity type.
Declared interests
The study was funded by the National Center for Medical Rehabilitation Research. No other conflicts of interest were declared in the provided text.
The easy way to misread this
Do not assume these models can accurately measure physical activity intensity or energy expenditure in daily life. They currently distinguish 'walking' from 'not walking' well, but cannot reliably tell if a child is walking comfortably or briskly, and have not been tested outside of a clinic.