Lameness scoring system for dairy cows using force plates and artificial intelligence.
Authors: Ghotoorlar S Mokaram, Ghamsari S Mehdi, Nowrouzian I, Ghotoorlar S Mokaram, Ghidary S Shiry
Journal: The Veterinary record
Summary
# Editorial Summary: Force Plate-Based Lameness Detection in Dairy Cattle Subjective visual lameness scoring remains the standard in dairy herd health screening, yet it suffers from poor repeatability both within and between observers, limiting its utility for longitudinal monitoring. Mokaram and colleagues developed an automated system using four balanced force plates integrated into a hoof-trimming box to capture ground reaction force (GRF) data from 105 dairy cows, extracting 23 kinetic variables that were processed through artificial neural networks to generate objective lameness scores comparable to conventional visual assessment. The automated system demonstrated superior repeatability to subjective methods (mean coefficient of variation 14.55%), with sensitivity and specificity both exceeding 72% across locomotion score groups 1–4, though performance declined for severely lame animals (group 5: 50% sensitivity, 100% specificity). For equine professionals accustomed to force plate gait analysis, this bovine application highlights the potential of objective kinetic assessment to standardise lameness detection across different species and contexts. Whilst the higher specificity in severe lameness may limit detection of the most compromised animals, the system's reliability and objectivity could substantially improve early-stage lameness identification and herd health management protocols, particularly where consistent screening is essential.
Read the full abstract on PubMed
Practical Takeaways
- •Automated lameness detection systems using force plates show promise as more objective and repeatable alternatives to subjective visual scoring in dairy herds
- •The system performs well for detecting non-lame and severely lame animals but has lower sensitivity for severe lameness (group 5), suggesting it may be most useful for screening rather than final diagnosis
- •High repeatability (CV 14.55%) suggests this technology could enable consistent longitudinal monitoring of individual animals' lameness status over time
Key Findings
- •Automated lameness scoring using force plates and artificial neural networks achieved mean coefficient of variation of 14.55% demonstrating high repeatability compared to subjective scoring
- •System demonstrated sensitivity and specificity higher than 72% across locomotion score groups 1-4, with 100% specificity and 50% sensitivity for group 5
- •Twenty-three features extracted from ground reaction force data were used to train the artificial neural network on 60% of data from 105 dairy cows