Quantitative lameness assessment in the horse based on upper body movement symmetry: The effect of different filtering techniques on the quantification of motion symmetry
Authors: F. S. Bragança, C. Roepstorff, M. Rhodin, T. Pfau, R. V. Weeren, L. Roepstorff
Journal: Biomed. Signal Process. Control.
Summary
# Editorial Summary Quantitative gait analysis systems are increasingly used to detect lameness in horses, yet different research groups and commercial platforms employ varying signal processing approaches, raising questions about the reliability and comparability of results across systems. Bragança and colleagues evaluated five filtering techniques commonly used in equine kinematic analysis—two infinite impulse response filters (Butterworth and Chebyshev), signal decomposition, and moving average methods—using both theoretical signals and data from horses with experimentally induced lameness, focusing particularly on how each method affected the calculation of upper body symmetry parameters. The Butterworth filter and signal decomposition method proved most effective at removing unwanted noise whilst preserving the genuine motion data needed for lameness detection, whereas improper selection of cut-off frequencies in IIR filters could lead to false negatives, with results falling outside predefined reference ranges for lameness identification. The authors emphasise that optimisation of filtering parameters through residual analysis is essential for accurate quantification, and their findings highlight that practitioners and researchers using commercial gait analysis systems should understand the filtering methodology underpinning their chosen platform to ensure valid interpretation of symmetry parameters. This work provides a technical foundation for standardising signal processing across equine gait analysis systems and helps explain why results may differ between laboratories or commercial devices.
Read the full abstract on the publisher's site
Practical Takeaways
- •When using quantitative gait analysis systems for lameness detection, understand which filtering method your system uses—different filters can give different results from the same horse
- •Be aware that poorly optimized filters may miss lameness or create false positives; if results seem inconsistent with clinical observation, filtering parameters may be the issue
- •When comparing lameness quantification data between different systems or research, filtering technique differences may explain apparent discrepancies in symmetry measurements
Key Findings
- •Improper selection of cut-off frequency for IIR filters can produce false negative results in lameness detection, misclassifying lame horses as sound
- •IIR Butterworth filter and signal decomposition method provided superior reduction of unwanted signal components compared to Chebyshev, moving average, and other techniques
- •Optimization of filtering techniques is critical for objective lameness quantification; different systems using different filter methods may produce inconsistent symmetry parameters from identical data
- •Residual analysis can be used to fine-tune filter settings to optimal parameters for each filtering technique