Abstract
Upright sitting is one of the first motor skills an infant learns, and thus sitting postural control provides an early window into the infant's motor development. Early identification of infants with motor developmental delay, such as infants with cerebral palsy, allows for early therapeutic intervention by physical therapists. Early intervention is thought to produce better outcomes, due to greater neural plasticity in younger infants. Postural sway, as measured by a force plate, can be used to objectively and quantitatively characterize infant motor control during sitting. Pathology, such as cerebral palsy, may alter the fractal properties of motor function. Often physiologic time series data, including infant sitting postural sway data, is mathematically non-stationary. Detrended Fluctuation Analysis (DFA) is useful to characterize the fractal nature of time series data because it is does not assume stationarity of the data. In this study we found that suitable selection of the order of the detrending function improves the performance of the DFA algorithm, with a higher order polynomial detrending better able to distinguish infant sitting posture time series data from Brown noise (random walk), and first order detrending better able to distinguish infants with motor delay (cerebral palsy) from infants with typical development.