Detecting Pain-Related Grooming in Rats Using Deep Learning
Author Information
Author(s): Lee Chien-Cheng, Lui Ping-Wing, Gao Wei-Wei, Gao Zhongjian, Cataldo Andrea, Danishvar Sebelan, Valencia Miguel
Primary Institution: Department of Electrical Engineering, Yuan Ze University, Taoyuan, Taiwan
Hypothesis
The proposed two-stage activity detection method can detect grooming episodes with a satisfactory level of accuracy and pain-related activity is indicated by an increased frequency of episodes within a specific range of grooming duration.
Conclusion
The study successfully developed a method to detect grooming behaviors in rats, showing that pain-related activity is reflected in longer grooming episodes.
Supporting Evidence
- The model achieved an average validation accuracy of 88.5%.
- Pain-related grooming behaviors were induced by injecting pain-inducing chemicals into the rats.
- The study proposed a new response metric to indicate pain-related grooming behavior.
- Statistically significant changes in grooming behavior were observed consistent with pain expression.
Takeaway
Scientists created a computer program to watch videos of rats and figure out when they are grooming themselves because of pain, helping to understand how animals show they are hurting.
Methodology
The study used a two-stage grooming activity detection scheme analyzing day-long video recordings of rats to identify grooming behaviors.
Potential Biases
Potential bias in human classification of grooming clips and variability in animal responses to pain inducers.
Limitations
The sample size is small, and human judgment of grooming behavior can be uncertain due to the use of single-camera images.
Participant Demographics
Male adult Sprague Dawley rats, body weight 250–300 gm.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website