Selection of Image Texture Analysis and Color Model in the Advanced Image Processing of Thermal Images of Horses following Exercise.
Authors: Domino Małgorzata, Borowska Marta, Kozłowska Natalia, Trojakowska Anna, Zdrojkowski Łukasz, Jasiński Tomasz, Smyth Graham, Maśko Małgorzata
Journal: Animals : an open access journal from MDPI
Summary
# Editorial Summary Infrared thermography (IRT) has potential as a non-invasive tool for monitoring exercise-induced physiological changes in horses, yet translating thermal data into clinically meaningful information requires robust image analysis protocols. Researchers examined twelve Polish warmbloods over six consecutive exercise tests, collecting thermal images and blood samples before and after each session, then subjected all 144 images to advanced texture analysis across four colour models (RGB, YUI, YIQ, and HSB) and eight texture-feature approaches encompassing 88 individual features. Only twelve texture features correlated with blood biomarkers of exercise effect, predominantly from the RGB colour model (nine features) with smaller contributions from YIQ and HSB, and these relationships were most consistent in the thoracolumbar region. Variance, sum of squares, and sum of variance calculations proved highly repeatable across processing protocols, suggesting these metrics offer reliable baseline measures for longitudinal monitoring. For equine professionals implementing IRT in practice—whether for performance assessment, recovery monitoring, or injury prevention—the findings advocate standardising image analysis to RGB or YIQ colour models combined with histogram statistics and gray-level co-occurrence matrix approaches, which should improve consistency and diagnostic utility across different thermal imaging systems.
Read the full abstract on PubMed
Practical Takeaways
- •Standardized thermal image processing using RGB or YIQ color models with histogram statistics or co-occurrence matrices may provide objective, non-invasive monitoring of exercise response in horses
- •The thoracolumbar region is the optimal anatomical location for thermal texture analysis to detect post-exercise physiological changes
- •This method could complement traditional blood sampling for monitoring training stress and recovery in performance horses
Key Findings
- •12 texture features across RGB (9), YIQ (1), and HSB (2) color models were significantly related to blood biomarkers following exercise
- •RGB color model variance, sum of squares, and sum of variance showed high repeatability between image processing protocols
- •Histogram statistics and gray-level co-occurrence matrix texture approaches combined with RGB and YIQ color models are optimal for equine infrared thermography analysis
- •Thoracolumbar region thermal image texture analysis can be used to assess blood biomarker responses to standardized exercise over six consecutive days