How GrimACE Measures Pain in Mice

Pain can be difficult to recognise, particularly in animals unable to tell us how they feel. That difficulty has prompted scientists to develop tools such as this one.

At first sight, the apparatus resembles an ordinary plastic container: white sides, a vivid orange floor and no elaborate features. It could easily be mistaken for something from a playroom.

Yet important research takes place inside. Scientists are using the box to learn how animals experience pain and how it might be reduced.

Designed for mice used in research, the box provides a quiet, dark space in which they can behave naturally.

There is no person hovering close by or making sudden movements. Instead, the controlled setting allows small details to be observed without adding stress.

Why darkness matters

The box is completely dark inside, and there is a reason for that. Mice feel more secure in darkness, and feeling safe helps them behave normally. They explore and sniff their surroundings before settling down; some even go to sleep.

Everything is recorded by two cameras, one positioned above the mouse and another in front. Infrared light allows them to see without disturbing the animal.

Although the equipment records slight changes in movement and posture, its main attention is on the face.

Pain can appear through subtle signs: a mouse may narrow its eyes, move its ears or alter the position of its whiskers. Such details are easily overlooked by the naked eye, particularly while the animal is moving.

How GrimACE reads pain in a face

For many years, scientists have used the Mouse Grimace Scale to estimate pain by scoring facial features. Each sign is assigned a number from 0 to 2 according to how clearly it appears. The approach is effective, but not without limitations.

Scoring takes time and can vary according to the person carrying it out. Sometimes, too, the mouse’s face cannot be seen clearly enough for a proper assessment.

A human observer’s presence can also cause the animal stress, potentially influencing the results.

The new system, known as GrimACE, takes a different approach: an algorithm examines the mouse’s face in real time.

Recording starts as soon as the animal enters the box. The system picks out the clearest frames and analyses them for indications of discomfort.

Improving how animals are used

The research comes from a team at ETH Zurich led by Oliver Sturman, Head of the 3R Hub.

The Hub aims to improve the use of animals in science by applying the principles of Replace, Reduce and Refine.

“This box allows laboratory animals to be observed in a humane and standardised way, whether by us here in Zurich or by researchers on the other side of the world,” said Sturman.

“This is important so that the animals feel comfortable and unobserved.”

A more consistent way to measure

Consistency is a major challenge in animal research. Two specialists may assess the same mouse and give it different scores. This is not a matter of carelessness: human judgement naturally varies.

During one test, three trained raters assessed identical images but did not give matching scores. Although each rater was consistent with their own assessments, their overall scoring levels varied.

“We secretly gave all three raters the same images to assess, to check whether their own scores were consistent,” said Sturman. Each person’s scores were consistent, but the differences between raters remained evident.

“This is where we see the strength of the computer, as it delivers standardised results.”

Such consistency is important. Underestimating pain could mean animals do not receive necessary care, while overestimating it could bring experiments to an early end.

“If someone always assesses that an animal is not in pain, animals will suffer needlessly. And if someone always gives overly high scores, there is a risk that experiments are abandoned unnecessarily,” said Sturman.

Tested in real conditions

The researchers tested GrimACE in a study of mice that had undergone brain surgery.

The mice were given painkillers in line with expert guidelines, and images were captured both before and after surgery.

An expert manually assessed thousands of the images while the algorithm examined the same data. Their results were almost identical.

The close agreement indicates that the system can match trained human observers, while working more quickly and without becoming tired.

More than just faces

Facial expressions are not the system’s only focus. From above, a high-resolution camera also tracks the animals’ body movements.

It measures the distances between body points, shifts in angles and the speed at which those points move.

These measurements can expose signs of discomfort that people might overlook. Machine learning models detect patterns too subtle for human eyes to discern.

Built for global use

The complete setup has been designed for sharing, with its hardware and software available as an open-source kit. Laboratories worldwide can therefore build the same system and gather comparable data.

“The idea is that as many users as possible can assemble and use it in a straightforward and standardised way – and that the data will then be comparable,” said Sturman.

The system improves as it receives more data. “The more people who use GrimACE, the less bias there will be.”

Scientists in countries including the United States and the United Kingdom have already expressed interest. GrimACE is now also available for wider use at the ETH Phenomics Center.

What comes next

The team intends to continue developing the technology. It is considering whether to make it a commercial product or keep sharing it freely through collaborations. For now, the researchers have a clear focus.

“We’re currently sharing our knowledge and technology in collaborations and are focusing on mutual data exchange to improve the system,” said Sturman. “Our primary concern is to improve animal welfare.”

The small plastic box may appear unremarkable, but inside it a quiet change is taking place in the way science treats the animals on which it relies.

The complete study was published in the journal Lab Animal.