The Mouse and the Machine

How video surveillance and machine learning improve aging studies

Every single human in the world faces the same reality: aging. According to the World Health Organization, by 2050, 22% of the world's population will be over 60. It is therefore necessary for aging research to be a top priority for doctors and researchers. And as usual when researchers look deeply into a biological question, our mouse models are right beside us.

Calico Life Sciences is one company that is looking deeply into the mechanisms of aging in order to improve the health and longevity of patients. By observing the frailty of aging mice (some of which are sourced from Charles River), they gain insight into age-related conditions in people. However, handling the mice to record their progression can be detrimental to the animal.

"One of the issues that we find with aging studies is that frailty scoring is time intensive and stressful for the animals," said Ellie Karlsson, Attending Veterinarian and Director at Calico. "Even if it's not invasive, it's not something that the animals want to experience, especially as they become more and more frail."

Removing a mouse from its home to test its gait, grip strength, or other frailty factors is stressful for them. Even with improved low stress handling techniques, it is best to reduce the need for handling at all. So the researchers at Calico came up with a new video monitoring system to watch the mice as they go about their normal day - eating, sleeping, running on their wheel, and interacting with their bedding material.

"We had this idea that could we put together a package that would replace manual frailty assessments," Karlsson said. "We designed an experiment around validating our digital measures versus our manual frailty measurements."

It was not as easy as substituting a 1:1 measurement from manual to video. The team had to come up with new frailty criteria based on the mouse's natural behavior. This proved beneficial not only for the mouse's well-being, but also produced richer, more flexible data.

"One of the advantages of the video-based system is that the metrics can be developed post hoc," said Graham Ruby, principal investigator at Calico. "Once the video is collected, we have a set of metrics already built and ready to be applied. But if we want to update our criteria laterwe can retroactively reevaluate frailty from a study."

Instead of having to repeat the same studies with new parameters, the previously recorded footage can simply be reviewed with new parameters. This replaces mice that would have been used for a whole new study and reduces the overall number of mice needed to continue their research.

There are other advantages for a machine learning-based video monitoring system. Circadian rhythm is known to change with age, but it would not be practical for a scientist to stare at the mice 24/7 to note when they are active. The camera, however, never gets tired. And even a simple computer program could tell the difference between a mouse that is awake or asleep, offering a valuable data point that was not feasible to collect in the past.

And Calico's machine learning algorithm is far from simple. The program, designed by Ruby and his team, can detect changes in the animal's coat; time, speed, and distance on the wheel; and even body weight fluctuations with excellent accuracy. The machine learning algorithm was also trained on a variety of mice, both inbred and outcross, to ensure the frailty scoring was as robust as possible. As researchers learn more about the physical manifestations of aging, they can apply new criteria to their observations, teaching the computer more about what to look for.

Freeing up researcher time, improving the stress levels of the mice, and capturing raw video data that can be analyzed later all fall within the goals Calico set for their aging research. According to Karlsson, the impetus for switching to video monitoring was a core value of embracing and utilizing advanced technologies. In this case, that has also led to better, more flexible science.

This is part of Eureka's ongoing series of stories focused on animal welfare.