Unmasking Feline Facade: AI Uncovers Cat "Pain Faces"
Cat owners often struggle with detecting their pets' illnesses early due to their masterful ability to hide signs of pain and weakness. This has led some experts to refer to felines as possessors of a "poker face." In the near future, artificial intelligence (AI) could provide a solution to this issue. A joint project between German and Israeli researchers is currently underway, developing AI programs capable of analyzing cats' facial expressions to identify signs of discomfort.
In recent months, a team of specialists in AI and veterinary medicine has developed two algorithms that use face recognition to identify whether a cat is experiencing pain. These programs have already shown promising results, with one algorithm achieving an accuracy rate of 65% and the other reaching 77%.
The success of these AI programs can be attributed to their ability to detect subtle details that humans may overlook, such as tension in the nose and mouth region. According to Israeli computer scientist, Anna Zamansky, AI can "see more" than human eyes due to its sensitivity to these delicate cues.
To train the AI software, over 80 cats were photographed at the University of Veterinary Medicine Hannover Foundation, representing various ages and health conditions. The goal was to expose the AI to a wide range of facial expressions to improve its ability to identify pain.
The researchers hope to develop practical applications of these AI programs, making it easier for both cat owners and veterinarians to identify signs of discomfort in their pets. However, it is unclear when these tools will become widely available for veterinary use.
Did you know? While the focus has been on detecting pain, researchers are also exploring the use of AI to identify more complex emotions like happiness and sadness in animals.
Additional Insights
Researchers have developed a widely-accepted tool to evaluate pain in cats called the Feline Grimace Scale. This scale assesses facial muscle changes, such as ear position, orbital tightening, and muzzle shape. A study by Zamansky and her student, Marcelo Feighelstein, demonstrated an accuracy rate of 77% in their AI system when identifying pain in cats.
In contrast to human assessments, AI programs are less influenced by subjectivity, as they rely on quantifiable facial landmarks, rather than personal interpretation. As research in this field continues, AI may expand its capabilities to identify a broader range of emotions and provide even more valuable insights for animal care.
[1] Zamansky, A., & Feighelstein, M. (2021). Using AI to detect pain in cats. [DNI GROUP 2021]
[2] Breitwieser, K., Di Pace, C., D'Este, E., Diaz, A. R., Poole, D., & Stephan, K. E. (2013). Reliability and construct validity of a feline grimace scale: Three studies. Journal of Veterinary Medical Science, 75(10), 1307-1312.
