Deep learning for video-based automated pain recognition in rabbits.

Area of Science:

• Veterinary medicine
• Animal welfare science
• Artificial intelligence in animal health

Background:

• Rabbits (Oryctolagus cuniculus) are vital research models and companion animals.
• Accurate pain assessment in rabbits is crucial but remains understudied.
• Current methods for rabbit pain evaluation are limited.

Purpose of the Study:

• To develop and validate an automated system for detecting acute postoperative pain in rabbits.
• To establish an objective and reliable method for pain recognition in Oryctolagus cuniculus.
• To improve the welfare of rabbits in research and pet settings through enhanced pain monitoring.

Main Methods:

• Utilized video footage of 28 rabbits, comparing pre-surgery (no pain) and post-surgery (pain) states.
• Developed an AI model analyzing both facial expressions and body posture for pain indicators.
• Employed Grayscale Short-Term stacking (GrayST) with 1-second interval sampling for temporal analysis.
• Implemented a frame selection technique to optimize video data utilization.

Main Results:

• The AI model achieved a pain recognition accuracy exceeding 87%.
• The combined analysis of facial and body posture data proved effective for pain detection.
• The GrayST method successfully incorporated temporal information for accurate video classification.

Conclusions:

• The developed AI model offers a promising automated solution for acute postoperative pain detection in rabbits.
• This technology can significantly enhance the objective assessment of pain in Oryctolagus cuniculus.
• Automated pain recognition systems can improve rabbit welfare in experimental and domestic environments.