Automated measurement of stress scores in video recordings of laboratory mice

点击翻译按钮获取中文摘要

Using laboratory mice for biomedical research requires strict monitoring of animal wellbeing. Especially, close monitoring of stress and pain is a key requirement for any pre-clinical study incorporating laboratory mice. Numerous protocols for mouse stress assessment have been developed including grimace analysis using the mouse grimace scale (MGS) or analysis of behavioral patterns such as burrowing and nesting behavior that all differ strongly between healthy and non-healthy individuals. While these methods are well established and tools such as MGS show a high correlation between visually acquired score and baseline measurements, assessing animal wellbeing with these scales is a time-consuming task. The animals need to be placed into special environments that are monitored using video equipment and the videos are subsequently analyzed manually. Furthermore, to ensure comparability and to reduce inter-rater variability, it is common practice to have the same expert analyze all videos of a trial.The aforementioned requirements and restrictions result in a high workload and planning overhead for the laboratory personnel. Therefore, this project aims at developing novel methods for automated behavioral analysis and assessment of pain and stress in laboratory mice. The developed algorithms will allow continuous, objective, quantitative and reproducible analysis of acquired video data. The primary focus will be on the development of methods supporting existing imaging setups or those being currently developed in parallel research projects, allowing quick deployment of the developed methods into existing lab environments and retrospective analysis of already existing video data.During the first project stage, a method automatically assessing of the mouse grim scale will be developed to allow robust automated analysis of animal grimacing. In the second stage, we aim at developing methods for robust analysis of behavioral experiments with laboratory mice filmed in open environments. These experiments include analyzing the nesting and burrowing behavior of mice, two experimental setups that are also currently being evaluated manually with considerable manual effort. Two other work packages include extending our methods to home cage analysis for continuous assessment and research on novel image-based and automatically detectable markers for data-driven pain assessment. We aim at developing methods that allow continuous real-time assessment of the animal state with minimal required user interaction, resulting in drastically improved response time, work effort, evaluation quantification and an overall contribution to the animals’ wellbeing.The developed methods will be made publicly available as open source software bundle allowing incorporating them into existing imaging setups.