Machine Learning-Guided Video Analysis Identifies Sound-Evoked Pain-Related Behaviors from Facial Grimace and Body Cues in Mice.

Humans can experience sound-evoked pain, either from extremely loud sounds or in cases of pain hyperacusis from typically tolerable sounds. However, the mechanisms underlying sound-evoked pain remain poorly understood. Developing behavioral methods to measure sound-evoked pain in animal models is critical for elucidating these mechanisms. Here, a machine learning-based approach was developed to measure sound-evoked pain in freely moving mice of both sexes by analyzing facial grimace and body position from video recordings during sound exposure. Facial grimace, a commonly used method to detect pain in mice, and body position, which can be used to measure postural and movement changes that also indicate pain, were both quantified using a deep neural network model trained to extract established facial and body features from video recorded by a single camera. To validate the model's capability to detect pain, a known painful state, migraine induced by the injection of the neuropeptide calcitonin gene-related peptide (CGRP), was used. Using this machine learning-based approach, the ability to quantify a pain response from CGRP-induced migraine, distinct from baseline behavior, was demonstrated, resulting in a defined pain threshold. Sound exposures at high intensities elicited significant changes in facial grimace and body position, in comparison, surpassing the pain threshold calculated from CGRP-induced migraine. These behavioral changes were absent in Tmie-knock-out mice, which lack functional sound transduction in the cochlea. This automated, high-throughput framework enables objective and sensitive analysis of pain-related behavior providing a foundation for future studies investigating the peripheral and central mechanisms of sound-evoked pain.