Statistical analysis of the 180 degree walking turn: Common patterns, repeatability and prediction bands of turn signals
Abstrakt
Turning is an essential movement and has been shown to be a relevant measure for differentiating pathologies. Nowadays, turn analyses utilizing inertial measurement units (IMU) have expanded.
Although several IMU-based turn metrics exist, there is no information on the repeatability of turn signals and on the existence of signal patterns shared across subjects. Also, the variability of IMU signals within various subject groups has not been estimated yet.
This paper presents an analysis of turn angular velocity and acceleration provided by IMU and tests them for repeatability, patterns, and variability within groups of healthy and diseased subjects. Intra-class correlation and methods for estimating prediction bands, namely the Gaussian point-by-point and bootstrap method, were employed to analyze turn signals from Parkinson disease patients and a control group.
The yaw angular velocity demonstrated the highest repeatability in both groups as well as reliability of a shared pattern (p = 0.79 and 0.86). The bootstrap method showed wider bands and higher true coverage in comparison to its Gaussian counterpart.
From the results of the performed analysis, we recommend the bootstrap method for determining prediction bands. We also recommend the yaw angular velocity as the signal to be assessed in turn analysis.