AI Enhances Mouse Behaviour Analysis, Reducing Animal Use in Research

ronaldfrenn

Researchers at ETH Zurich are using artificial intelligence to improve the efficiency of analyzing laboratory mice behaviour and reduce the number of animals needed in experiments.

Assessing the wellbeing of animals in research is a crucial skill for stress researchers and those working to improve animal welfare. Since animals cannot directly communicate their feelings, researchers must rely on behavioural observations to gauge their condition. To advance this process, researchers led by Johannes Bohacek, Professor at the Institute for Neuroscience at ETH Zurich, have developed a new method that significantly enhances the analysis of mouse behaviour.

The method uses automated behavioural analysis powered by machine vision and artificial intelligence. In this process, mice are filmed, and their behaviour is automatically analyzed from the video recordings. Traditionally, analysing animal behaviour required painstaking manual work, often taking days to complete. However, many leading laboratories have transitioned to more efficient automated behavioural analysis systems in recent years.

One challenge with this approach is the vast amount of data generated. The larger the dataset and the more subtle the behavioural differences, the greater the risk of being misled by artefacts. For example, automated processes might mistakenly classify irrelevant behaviours as significant. The typical solution to this problem in research is to increase the number of animals tested to account for such artefacts and still produce reliable results.

The new method developed by ETH researchers allows for meaningful results and the identification of subtle behavioural differences even with a smaller group of animals. This not only helps reduce the number of animals used but also enhances the quality of each experiment. The method aligns with the 3Rs principle—Replace, Reduce, and Refine—adopted by ETH Zurich and other institutions, aiming to replace or reduce animal experiments and refine methods to improve their efficiency.

The ETH method focuses on both the specific patterns of behaviour in mice and the transitions between these behaviours. Typical behavioural patterns include standing on their hind legs when curious, staying close to walls when cautious, and exploring new objects when confident. Even when a mouse is still, its behaviour can be informative, suggesting alertness or uncertainty.

The transitions between these behaviours are also significant. For example, rapid switching between behaviours may indicate stress, while stable, less abrupt transitions suggest relaxation. The ETH researchers' method combines these transitions mathematically into a single value, which strengthens the statistical analysis.

Professor Bohacek, a neuroscientist and stress researcher, is using these insights to explore how mice respond to stress and how certain brain mechanisms influence an animal’s ability to cope. The hope is that by understanding these processes, researchers can predict how animals handle stress and potentially develop therapies for human populations at risk.

The new method has already been successfully used to study mice's responses to stress and medication. The statistical improvements have allowed the team to detect subtle behavioural changes, such as the different ways acute and chronic stress affect mice. These behavioural shifts are linked to varying brain mechanisms, advancing the understanding of stress responses in animals and humans.