A methodological framework to analyze age-related biomechanical differences in running using a single IMU worn on the lower back

Recreational running is a popular activity worldwide, but biomechanics studies have traditionally relied on lab-based treadmills and cameras, limiting analysis to controlled environments and few consecutive strides. Despite the rise of wearables for outdoor data collection, methodological guidelines for IMU-based overground running analysis remain scarce, particularly regarding signal filtering and data volume required. In addition, applications beyond acceleration quantification to real-world problems such as age-related differences are also scarce. 66 healthy runners completed two 400-m overground runs at two different speeds. A single IMU (256 Hz) was placed at L5. Raw accelerations were filtered using ten 4th-order low-pass Butterworth filters: four from literature cut-offs and six from residual analysis and 95% power spectral densities. Descriptive stats and repeated-measures ANOVA evaluated condition differences. Nonlinear metrics (α from DFA, Higuchi fractal dimension, sample entropy) were assessed on a secondary ultrarunning dataset up to 800 strides, using cumulative means to examine metric behaviour with stride count. Age effects compared younger (18–25) and Masters (40+) groups on spatiotemporal parameters, mean acceleration peaks, and nonlinear metrics. Findings indicate that applying overly aggressive low-pass filters (<25 Hz) is problematic, as they can significantly alter metrics. Nonlinear metrics typically need long series; however, representative estimates of stride time variability and fractal properties are achievable from moderate datasets via sacral IMU (<250 strides). Age-related differences are subtle and mainly evident in temporal metrics, while complexity remains unchanged over 400 m. This dissertation highlights the theoretical and practical impact of different methodological frameworks on IMU-derived running metrics and applies them to a novel analysis of age-related real-world behaviour, providing a foundation for future research.