Variability of the EMG mean power frequency: A study on the trapezius muscle

Calculation of the EMG mean power frequency (MPF) is a common procedure applied in evaluation of the frequency shift associated with local muscle fatigue. Variations of the MPF that are unrelated to muscle fatigue may jeopardize the estimation of the frequency shift. Different kinds of variation include random variation and systematic variation due to changes in posture or load. In a previous article we have evaluated the systematic linear variation of the MPF. The aim of the present study was to examine the random variation. Data sequences of 10 s, each obtained from nonfatigued trapezius muscle of 19 healthy subjects, were examined over a functional range of load and joint angles with multiple regression analysis. The random variation was evaluated with residual analysis. The residual standard deviation within the whole group was 10% for surface recordings and 13% for intramuscular recordings. If only within-subject variation was considered, the corresponding values were 5 and 8%. Based on this, confidence and prediction intervals for the regression models were calculated. Ninety-five percent confidence intervals were ±1-3% around the regression surfaces, whereas 95% prediction intervals for single measurements were as large as ±20-26% for the whole group, and ±11-20% if only within-subject variations were considered. Assessment of localized muscle fatigue using single MPF estimates should therefore be avoided. Multiple measurements and regression analysis are discussed as methods to minimize the effects of random variations.