Running Gait Signal Filtering Strategies And Justified Cutoff Frequencies Are Imperative For Comparative Analytics

Ground reaction force (GRF) analyses guide clinical running gait interpretation and inform injury prevention and rehabilitation. Conventional gait analysis uses specific load rate region of interest (ROI), and signal filtering to attenuate high-frequency noise and artifacts. GRF waveform fluctuations with periods = 100th/s are expected vibrations from impact mechanics, not signal noise. Setting cutoff frequencies ~50 Hz smooths or removes these features and affects load rate. These strategies may eliminate important biomechanical phenomena needed for precise analytics. PURPOSE: To determine signal filtering effects using different cutoff frequencies on net GRF responses and load-rates in running gait. METHODS: Runners (N = 426; rearfoot strikers) ran on an instrumented treadmill; GRFs were sampled at 1200 Hz. Net GRF were post-processed across lowpass filters (4th-degree Butterworth filter; 600, 100, 60 & 10 Hz). Numerical time derivative of GRF provided the load-rate and characterized waveform shape. RESULTS: Signal filtering yielded differences in GRF rate change (Fig. A, B). Compared to 600 Hz, 100 Hz showed good fidelity; 60 & 10 Hz reduced impact load rate by 39-84% (p < .05) and peak GRF by 3-9% (p < .05), delayed time to initial impact peak by 5-93 ms (p < .05) or eliminated it, and smoothed vibrations. Waveform shape and vibration frequency varied across runners, but filtering effects on GRF waveforms were similar. 40% of max positive load rates occurred outside the ROI. Early (<30% stance) max negative load rates excluded by ROI reveal dramatic unloading before full loading. CONCLUSION: Lowpass filtering fundamentally alters GRF waveform structure, load rate and wave peak sharpness. Analytic capabilities are impaired by suboptimal signal filtering processes. Careful selection of filtering method is therefore vital for precise clinical interpretation and translation to injury prediction.