AI Powered GrimACE System Standardizes Laboratory Animal Welfare by Detecting Subtle Facial Expressions of Pain in Mice

ETH Zurich researchers launch GrimACE, an AI tool that uses facial recognition to provide objective, real-time pain assessment for laboratory animals.

By: AXL Media

Published: Apr 27, 2026, 6:01 AM EDT

Source: Information for this report was sourced from EurekAlert!

The Technological Pursuit of Humane Research

The ethical landscape of laboratory research is undergoing a digital transformation with the introduction of automated welfare monitoring. Oliver Sturman, Head of the 3R Hub at ETH Zurich, has introduced a standardized observation box designed to minimize animal suffering through non-invasive surveillance. This specialized environment, featuring black acrylic walls to ensure mice feel unobserved and comfortable, utilizes two cameras to capture behavior in total darkness. The goal is to refine the "3R" principles—Replace, Reduce, Refine—by providing a consistent method for researchers worldwide to assess animal well-being without the stress of human presence.

Cracking the Code of Rodent Facial Expressions

Pain in rodents is often revealed through subtle shifts in facial geometry that are difficult for the human eye to track consistently. The GrimACE system focuses on specific markers such as the narrowing of eyes, bulging of the nose and cheeks, and changes in ear or whisker positioning. While the Mouse Grimace Scale has existed as a manual tool, it relies on researchers comparing live animals to pictorial charts, a process that is frequently subjective. By applying an algorithm to real-time video, the new system can identify these micro-expressions instantaneously, allowing for faster medical intervention or adjustments to pain relief protocols.

Overcoming the Subjectivity of Human Observation

A critical study recently published in the journal Lab Animal highlighted the inconsistencies inherent in human welfare ratings. When three different experts were asked to assess the same images of mice following brain surgery, their scores varied significantly based on individual bias, even though each person remained consistent with their own internal logic. One rater tended to provide higher scores across the board, while another consistently trended lower. The GrimACE system eliminates this variability by delivering standardized results, ensuring that experiments are not abandoned unnecessarily due to overestimation of pain, nor are animals left to suffer due to underestimation.