Can a shoe-mounted IMU identify the effects of orthotics in ways comparable to gait laboratory measurements?

Abstract

Background: Footwear and orthotic research has traditionally been conducted within laboratories. With increasing prevalence of wearable sensors for foot and ankle biomechanics measurement, transitioning experiments into the real-world is realistic. However wearable systems must effectively detect the direction and magnitude of response to interventions to be considered for future usage.

Methods: RunScribe IMU was used simultaneously with motion capture, accelerometers, and force plates during straight-line walking. Three orthotics (A, B, C) were used to change lower limb biomechanics from a control (SHOE) including: Ground reaction force (GRF) loading rate (A), pronation excursion (A and B), maximum pronation velocity (A and B), and impact shock (C) to test whether RunScribe detected effects consistent with laboratory measurements. Sensitivity was evaluated by assessing: 1. Significant differences (t-test) and effect sizes (Cohen's d) between measurement systems for the same orthotic, 2. Statistical significance (t-test and ANOVA) and effect size (Cohen's d & f) for orthotic effect across measurement systems 3. Direction of orthotic effect across measurement systems.

Results: GRF loading rate (SHOE: p = 0.138 d = 0.403, A: p = 0.541 d = 0.165), impact shock (SHOE: p = 0.177 d = 0.405, C: p = 0.668 d = 0.132), pronation excursion (A: p = 0.623 d = 0.10, B: p = 0.986 d = 0.00) did not significantly differ between measurement systems with low effect size. Significant differences and high effect sizes existed between systems in the control condition for pronation excursion (p = 0.005 d = 0.68), and all conditions for pronation velocity (SHOE: p < 0.001 d = 1.24, A: p = 0.001 p = 1.21, B: p = 0.050 d = 0.64). RunScribe (RS) and Laboratory (LM) recorded the same significant effect of orthotic but inconsistent effect sizes for GRF loading rate (LM: p = 0.020 d = 0.54, RS: p = 0.042 d = 0.27), pronation excursion (LM: p < 0.001 f = 0.31, RS: p = 0.042 f = 0.15), and non-significant effect of orthotic for impact shock (LM: p = 0.182 d = 0.08, RS: p = 0.457 d = 0.24). Statistical significance was different between systems for effect of orthotic on pronation velocity (LM: p = 0.010 f = 0.18, RS: p = 0.093 f = 0.25). RunScribe and Laboratory agreed on the direction of change of the biomechanics variables for 69% (GRF loading rate), 40%-70% (pronation excursion), 47%-65% (pronation velocity), and 58% (impact shock) of participants.

Conclusion: The RunScribe shows sensitivity to orthotic effect consistent with the laboratory at the group level for GRF loading rate, pronation excursion, and impact shock during walking. There were however large discrepancies between measurements in individuals. Application of the RunScribe for group analysis may be appropriate, however implementation of RunScribe for individual assessment and those including pronation may lead to erroneous interpretation.