Automated Gait Variability Assessment In Real-World Running Using Wearable Accelerometry

Abstract

Using wearable sensors to assess running gait variability may be a valuable tool to identify deterioration in performance or health in competitive and recreational runners. PURPOSE: To measure associations between gait variability and training intensity, mood state, and perceived fatigue in collegiate runners (CR) and novice/recreational runners (NR). METHODS: 30 CR wore a waist-mounted ActiGraph GT3X+ accelerometer for all non-interval and non-competition training sessions for a full season. 16 NR wore a waist-mounted GT3X+ and an on-shoe activity monitor (MilestonePod) during 13 weeks of progressive training. Subjects completed written (CR) or email (NR) daily surveys. Recently developed running and step recognition algorithms were applied to isolate running gait cycles from raw sub-second level accelerometer data. Gait variability was assessed using amplitude deviation of resultant acceleration. Associations between variability and training intensity, mood state, and fatigue were assessed with generalized additive mixed models. RESULTS: 1069 runs and 7.64 million steps were analyzed. Variability was greater in CR than NR (0.301 vs 0.262g, p=0.019). Within individual runs, variability was stable from 10-60 minutes in both CR and NR (Fig. 1). In CR, mood state was not significantly associated with variability (p=0.70). In NR, variability increased nonlinearly with average running speed (p<0.001), and a 1-point increase in fatigue on a 0-10 scale was associated with a 0.003g decrease in variability (p=0.051). CONCLUSION: A runner’s gait variability may be more strongly influenced by skill level and running speed than by acute changes in perceived fatigue or mood state. Unmeasured variables, such as training surface, may also affect gait variability. Figure 1. Gait variability in CR and NR across different individual runs (top left, black dots), across different levels of fatigue (top right), and within any given run (bottom left and right, thin lines).