Development of PainFace software to simplify, standardize, and scale up mouse grimace analyses

Area of Science:

• Animal behavior
• Computational biology
• Pain research

Background:

• Facial grimacing is a reliable indicator of pain in mammals.
• Current methods for scoring facial grimaces in mice rely on subjective human interpretation, leading to variability.
• Objective and automated methods are needed to standardize pain assessment in preclinical research.

Purpose of the Study:

• To develop and validate an automated, machine learning-based software platform, PainFace, for quantifying mouse facial grimaces.
• To assess the accuracy and reproducibility of PainFace across different laboratories and pain-inducing conditions.
• To enable high-throughput and objective measurement of spontaneous pain in mice.

Main Methods:

• Development of a cloud-based software platform (PainFace) utilizing machine learning algorithms.
• Detection of four key facial action units (orbitals, nose, ears, whiskers) associated with mouse grimace scale.
• Validation of PainFace accuracy using video recordings from mice undergoing laparotomy, carrageenan injection, and formalin injection in two independent laboratories.

Main Results:

• PainFace accurately scores facial grimaces in black-coated C57BL/6 mice on a 0-8 scale.
• The platform can generate grimace scores at a rate of up to 1 per second from standard video.
• Mice subjected to laparotomy surgery exhibited a 7-fold increase in time spent in a high grimace state compared to sham controls.

Conclusions:

• PainFace provides a reproducible and scalable method for quantifying facial grimaces indicative of spontaneous pain in mice.
• The automated platform enhances objectivity and reduces inter-observer variability in pain assessment.
• PainFace facilitates standardized and high-throughput analysis of facial grimace in preclinical pain research.