Researchers at ETH Zurich are leveraging artificial intelligence to analyze laboratory mouse behavior more efficiently, thereby reducing the number of animals used in experiments.
Researchers conducting animal experiments should have a high level of expertise in the statistical analysis of data. Researchers seeking to optimise the environment in which laboratory animals are preserved? To gauge the welfare of their animals, individuals desire to rely heavily on behavioral observations, since direct queries are not feasible, akin to assessing human wellbeing through non-verbal cues. Scientists from ETH Zurich’s Institute for Neuroscience, led by Professor Johannes Bohacek, have made significant strides in assessing mouse behavior, unveiling a groundbreaking approach that propels their research forward.
Using advanced computer vision and artificial intelligence technologies, the method enables automated behavioral evaluations. Researchers study mouse behavior by capturing and analyzing video footage using automated methods. While traditional methods for analyzing animal behavior once required tedious manual reviews, even commonplace today in most research settings, pioneering institutions have recently adopted innovative, eco-friendly automated behavioral assessment techniques.
The proliferation of information poses a significant challenge in its own right: the sheer volume of knowledge can overwhelm us? As the complexity of available data and refinements in behavioral patterns increase, so does the likelihood of being deceived by artifacts. These instances may inadvertently perpetuate an automated process for categorizing behaviors that are unrelated in the first place. Statistics suggests that increasing the number of animals studied could help mitigate the effects of artefacts and ultimately yield meaningful results.
Researchers at ETH have developed a novel approach that enables the detection of notable outcomes and nuanced behavioral differences among animals using smaller groups, ultimately reducing the number of animals required for experimentation and enhancing the significance of individual studies. Subsequently, this effort supports the three Rs initiatives undertaken by ETH Zurich and other research institutions. The 3Rs, a framework for refining animal experimentation, actually stands for replacement, reduction, and refinement, encouraging the adoption of alternative methods, minimizing the number of animals used, and optimizing experimental procedures.
The ETH researchers’ technique doesn’t merely utilize remote, highly specific patterns of animal behavior; instead, it concentrates intensely on the transitions between distinct behaviors.
Among the characteristic behavioral patterns exhibited by mice are standing upright on their hind legs when exhibiting curiosity, remaining near the periphery of their enclosure when cautious, and investigating novel objects with a sense of boldness. A mouse, stationary yet attentive, conveys a subtle dichotomy: poised between vigilance and uncertainty.
The fluidity of pattern transitions holds immense significance – animals exhibiting rapid, consistent shifts between distinct patterns may equally convey a sense of nervousness, emotional weight, or tension. In contrast, animals exhibiting calm or confident demeanors tend to display consistent behavioral patterns that are less likely to shift precipitously. These transitions are advanced. To streamline these calculations, the strategic approach mathematically aggregates their effects into a unified, substantial value, thereby fortifying statistical analyses with increased robustness.
ETH Professor Mark Bohacek is a renowned neuroscientist and leading expert in the field of stress research. He is exploring the neural mechanisms that determine an animal’s ability to cope with anxiety, examining which mental processes contribute to its effectiveness in managing anxious situations. “If behavioral analyses enable us to determine or predict an individual’s resilience in managing stress, we can investigate the exact neural mechanisms at play,” he notes. Potential remedies for sure-fire human danger teams can be derived from these analyses.
Using a cutting-edge approach, the ETH team has successfully decoded how mice respond to stress and specific medications in controlled animal studies. Due to advances in statistical analysis, even subtle differences between individual animal species can be distinguished and identified with precision. Researchers have demonstrated that acute stress and prolonged stress exposure significantly alter the behavior of mice in multiple ways. These connections between adjustments are intricately linked to distinct cognitive processes within the brain.
This innovative approach will seamlessly enhance standardization of tests, thereby enabling more accurate comparisons of results from diverse experiments, including those conducted by separate research groups.
“When employing synthetic intelligence and machine learning in behavioral evaluations, we’re advancing morally and sustainably the field of biomedical research,” states Bohacek. With expertise honed over several years, he has spearheaded efforts to explore the intricacies of 3R analysis alongside his dedicated team. The 3R Hub has been established at ETH to achieve this goal. The Hub aims to have a positive and constructive impact on animal welfare within the context of biomedical research.
The groundbreaking ETH 3R Hub technique has achieved its inaugural major triumph. According to Oliver Sturman, Head of the Hub and co-author of this study, they are delighted with the outcome. The 3R Hub facilitates the dissemination of innovative techniques to researchers within ETH and beyond, fostering collaboration and knowledge sharing. “Highly complex analyses such as ours necessitate extensive expertise,” Bohacek clarifies. “Overcoming the challenge of adopting innovative 3Rs methodologies is a major obstacle for many analytical laboratory settings.”
