Can grimace scales estimate the pain status in horses and mice? A statistical approach to identify a classifier.
Authors: Dalla Costa Emanuela, Pascuzzo Riccardo, Leach Matthew C, Dai Francesca, Lebelt Dirk, Vantini Simone, Minero Michela
Journal: PloS one
Summary
# Editorial Summary: Grimace Scales as Quantifiable Pain Assessment Tools in Horses Pain assessment in equine practice remains challenging, relying heavily on behavioural observation and performance changes rather than objective measures. Dalla Costa and colleagues applied rigorous statistical modelling to validate grimace scales—systematic scoring systems based on facial action units (FAUs)—as reliable pain classifiers in both horses and mice, testing whether specific facial movements could accurately identify animals in pain. Using advanced statistical techniques including linear discriminant analysis and support vector machines, the researchers analysed facial expression datasets and determined optimal weightings for individual FAUs, achieving greater than 70% accuracy in correctly classifying pain status. Their findings confirm that facial grimacing is a genuine pain indicator rather than coincidental facial movement, and identified specific FAU combinations most predictive of pain states in horses. For equine practitioners, these results provide evidence-based support for incorporating systematic facial expression assessment into pain evaluation protocols, whilst the proposed statistical framework opens the door to developing automated image analysis systems that could eventually offer real-time, objective pain detection—particularly valuable in situations where behavioural signs are subtle or confounded by other factors such as training stress or environmental novelty.
Read the full abstract on PubMed
Practical Takeaways
- •Grimace scales provide a reliable, objective method for assessing pain in horses based on facial expressions—useful when other pain indicators are unclear
- •The identified FAU weightings could enable development of smartphone/automated camera systems for real-time pain monitoring in stables, reducing reliance on subjective assessment
- •Facial expression scoring is species-specific; use HGS for equine patients and don't assume cross-species facial indicators translate directly
Key Findings
- •Facial Action Units in both Horse Grimace Scale (HGS) and Mouse Grimace Scale (MGS) are statistically related to actual pain status in animals
- •Statistical classifier achieved >70% accuracy in identifying animals in pain using optimal FAU weightings
- •Inter-rater reliability confirmed for both HGS and MGS, validating these scales as pain assessment tools
- •Individual FAU scores can be mathematically weighted and combined to create automated pain detection algorithms