It is a task many researchers would cede to automation and now, said Thomas Serre, assistant professor of cognitive, linguistic and psychological sciences at Brown University, it is finally possible.
Serre and colleagues from the Massachusetts Institute of Technology and California Institute of Technology have developed a computer program that is as accurate as people in identifying mouse behaviours in videos.
The team is making the fully customisable open-source software available for free. Given standard camcorder footage of a mouse, the software will automatically identify a mouse’s behaviour frame by frame.
The value of the software is that it could relieve graduate students and lab technicians from the boredom of performing similar tasks. It takes about 25 person hours to fully annotate an hour of mouse movies, so a small experiment with 10 mice who are each observed for five hours equates to 1,250 hours of work.
Because it is computerised, the system might also provide less subjective annotations than a human team would and could be less susceptible to bias.
There are a few commercial programs on the market that perform a similar sort of function, but some of them cost thousands of dollars, Serre said. They mostly base their behavioural coding on sensors, rather than video, and therefore have agreement rates with human observers of around 60 per cent − substantially lower than the greater-than-70-per-cent rate achieved by the new system.
Serre said that, at least for mice, the software is essentially ready to go for use in lab experiments and is likely to be easy to train to work with other laboratory animals.
UK not prepared for climate impacts, says CCC
Perhaps a Longtitude prize to solve railway line problems. "extreme heat causing further disruption through rail buckling and power line...