DEVELOPMENT AND APPLICATIONS OF AUTOMATED GRIMACE SCALES FOR THE STUDY OF PAIN AFFECT IN RODENTS

Emotional/affective states in mammals can be communicated to con- and allospecifics and through auditory and/or visual stimuli. The latter, visual stimuli, can present as changes in facial expressions denoting both the valence (negative or positive) and magnitude of the emotional state. For the study of negative affective states such as pain, grimace scales have proved to be a robust resource for measuring pain affect across species (humans, horses, seals, rodents, etc.), however, remain time and resource intensive measures. Here, we report the development of automated grimace scales that leverages computer vision/machine learning, hardware and software engineering, and cloud-computing to reliably detect and quantify the severity of grimace in laboratory mice. These automated grimace scales were developed using 100,000+ high-quality images manually scored by human experts across multiple preclinical mouse models and under diverse recording conditions, leading to a highly generalizable platform amenable for widespread adoption that also performs on par with human expert scorers. We leveraged this platform, coined PainFace, to report the behavioral underpinnings of analgesia and the translational potential of grimace as measure of pain. These findings lay foundational tools, analytical frameworks, and biological insights for the study of pain affect in rodents and have the potential to improve the modeling pain in preclinical subjects more effectively.