Fig 1.
Top: comparison of sample entropy values derived from tasks performed on the stability trainer (ST) or floor (Fl) in the medio-lateral (ML) plane (left) and anterior-posterior (AP) plane (right). This comparison is shown for m = 2 and m = 3 across 6 values of r: 0.05, 0.1, 0.15, 0.2, 0.25, 0.3. Sample entropy was consistently higher on the floor as compared to ST in both planes and for all m’s and r’s. Bottom: A between-group comparison of sample entropy for tasks performed on the floor. Across m’s and r’s, patients were consistently higher in the ML plane with no main effect of group but a significant interaction group × surface for r = 0.05 and r = 0.1. There were no between-group differences on the AP plane.
Fig 2.
Top: comparison of sample entropy values derived from tasks performed on the stability trainer (ST) or floor (Fl) in the medio-lateral (ML) plane (left) and anterior-posterior (AP) plane (right). This comparison is shown for m = 2 and r = 0.1 across different levels of down-sampling (DS) by 1, 2, 3 and 4 or when applying a 4th order low-pass Butterworth filter with a cutoff frequency at 10 Hz. Sample entropy was significantly higher on the floor for all DS levels in both planes but not when applying the filter. Bottom: A between-group comparison of sample entropy for tasks performed on the floor given DS or filter in the ML (left side) or AP (right side) plane.
Fig 3.
Sample entropy between surfaces (top) and between groups on the floor (bottom) in the ML direction (left hand side) and AP direction (right hand side) for original data, detrended data and differenced data when m = 2 and r = 0.1.
Fig 4.
We used the raw anterior-posterior time series (top) from one scene (the park scene, 120 seconds long) performed by a patient when standing on the floor to demonstrate the effect of detrending (middle) and differencing (bottom). The corresponding sample entropy values in this scene are: Raw: 0.463, Detrended: 0.56, Differenced: 2.6.
Fig 5.
Distributions of SampEn and e−SampEn as a function of the radius of similarity r for the data samples of patients and controls on the floor and stability trainer.
Fig 6.
On the left: distribution of SampEn for m = 2 and varying r on (X1,…XN) for N = 1000 where Xi′s are i.i.d. standard Gaussians; the mean (blue curve) is On the right: the function h(r) and its behavior for large r (on bottom: in log-scale).