Researchers are implementing artificial intelligence to better assess the behavior of lab mice, leading to enhanced efficiency and a decrease in the number of animals used in experiments.
Researchers at ETH Zurich are utilizing artificial intelligence to more effectively analyze lab mice behavior and minimize the number of animals in experiments.
One key skill that researchers focusing on animal welfare must possess is the ability to evaluate the wellbeing of their subjects through behavioral observations. Unlike humans, researchers cannot simply ask animals about their feelings. A team led by Johannes Bohacek, a Professor at the Institute for Neuroscience at ETH Zurich, has developed a new method that greatly enhances the analysis of mouse behavior.
This new approach uses automated behavioral analysis powered by machine vision and artificial intelligence. Mice are recorded on video, and these recordings are then analyzed automatically. Previously, scientists spent countless days manually assessing animal behavior—an approach still prevalent in many labs today. However, top laboratories have begun adopting more efficient automated methods in recent years.
Solving the Data Challenge
A challenge that arises from these new methods is the overwhelming amount of data generated. With more data and subtle behavioral differences to observe, the chance of misinterpretation increases. For instance, an automated tool may mistakenly classify an insignificant behavior as significant. Statistically, the solution to this problem is simple: test more animals to offset any inaccuracies and still achieve valuable insights.
The latest technique from ETH researchers allows investigators to derive meaningful insights and detect subtle behavioral variations among a smaller number of subjects. This advancement contributes to reducing animal use in research and enhances the relevance of individual animal experiments. This aligns with the 3R principles advocated by ETH Zurich and other institutions, which aim to replace, reduce, and refine animal testing through alternative methods and technological improvements.
Focus on Behavioral Consistency
The ETH team’s approach analyzes the specific, detailed patterns of mouse behavior while closely monitoring the transitions from one behavior to another.
Typical mouse behaviors include standing on their hind legs when curious, sticking close to cage walls when cautious, and showing curiosity about new objects when feeling brave. Even when a mouse is stationary, it conveys valuable information; it might be particularly aware of its surroundings or indecisive.
The shifts between these behaviors are significant—an animal that rapidly fluctuates between certain actions may be anxious or stressed. In contrast, a calm mouse tends to exhibit stable behavioral patterns and transitions less abruptly. To streamline this analysis, the method mathematically condenses these behavioral shifts into a single, meaningful metric, enhancing the robustness of statistical evaluations.
Enhanced Study Comparability
Professor Bohacek, a neuroscientist and expert in stress, is exploring which brain processes affect an animal’s ability to cope with stress. “If we can use behavioral studies to identify, or ideally, predict how well an individual manages stress, we can examine the specific brain mechanisms involved,” he states. Insights from this could lead to potential therapeutic options for certain vulnerable human populations.
The ETH team has already utilized this new method to investigate mouse responses to stress and various medications during experiments. With advanced statistical analysis, they can detect even minute differences between individual mice. For instance, they have distinguished how acute and chronic stress affect mouse behavior differently, which correlates with various brain mechanisms.
This novel approach also standardizes tests, thereby improving the comparability of results from different experimental settings, including those from various research teams.
Advancing Animal Welfare in Research
“By applying artificial intelligence and machine learning to behavioral assessments, we are enhancing the ethics and efficacy of biomedical research,” notes Bohacek. He and his team have been focused on 3R research for several years and established the 3R Hub at ETH to positively impact animal welfare in biomedical research.
“The new method represents a significant achievement for the ETH 3R Hub, and we are proud of it,” states Oliver Sturman, the Hub’s Head and co-author of this study. The 3R Hub is committed to facilitating access to this new method for other researchers at ETH and beyond. According to Bohacek, “Our analyses are intricate and require deep expertise. Many laboratories find it challenging to adopt new 3R approaches.” The aim of the 3R Hub is to support the implementation of these strategies to enhance animal welfare through practical assistance.
