Initial contact and toe off event identification for rearfoot and non-rearfoot strike pattern treadmill running at...

The purpose of this study was to determine the validity of kinematic based initial contact (IC) and toe-off (TO) identification algorithms for rearfoot and non-rearfoot runners across a broad range of treadmill running speeds. 14 healthy active participants completed six 20-60 s treadmill running trials at 6 speeds: 2.24, 2.68, 3.13, 3.58, 4.02, and 4.48 ms^-1. 3D kinematic data were collected for the last 20 s of each trial. Force plates (FP) were used as the gold standard to determine ICFP and TOFP for each step. Three algorithms for finding IC, IC_Milner, IC_Alvim, IC_Alvim-mod, and one algorithm for finding toe off, TO_Fellin, were chosen for analysis. Root mean square errors (RMSE) and difference scores with 95% confidence intervals were computed for IC, TO and stance time (ST). IC_Alvim RMSE ranged from 0.175 to 0.219 s. ST_Alvim RMSE ranged from 0.168 to 0.216 s. IC_Alvim-mod RMSE ranged from 0.105 to 0.131 s. ST_Alvim-mod RMSE ranged from 0.108 to 0.129 s. IC_Milner RMSE ranged 0.012 to 0.015 s. ST_Milner RMSE ranged 0.019 to 0.024 s. IC_Milner accuracy was inversely related to speed. IC_Milner corrected with a linear regression equation reduced differences to- 0.006 ± 0.012 s with 86% of foot strikes identified within 20 ms and 58% with 10 ms. TO_Fellin RMSE ranged from 0.012 to 0.016 s. IC_Milner adjusted for speed and TO_Fellin can be used to predict IC and TO within a broad range of treadmill running speeds (2.24-4.48 ms^-1) and for rearfoot and non-rearfoot strikers.