Enhanced health evaluation in mice using continuous home-cage monitoring and machine learning: a multicentric study

Ensuring the health of laboratory rodents is critical for ethical research and maintaining scientific integrity. Traditional daily visual observations by trained technicians, conducted during the rodents' sleep period, often fail to detect subtle but critical health indicators due to the short duration of inspections and obstructions from enrichment materials. Here we aimed to improve health checks in mice by utilizing continuous home-cage monitoring coupled with machine learning (ML) algorithms. We hypothesized that reduced locomotion in mice would indicate distress or sickness, and that continuous tracking would identify clinical cases earlier than visual checks. We retrospectively analyzed locomotion data from three institutions using the same sensor technology and applied ML/artificial intelligence (AI) models to generate digital alerts for potential clinical cases. These alerts were then compared with clinical records to verify the accuracy of the predictions. Our results demonstrated that the ML algorithm identified animals in distress -3 to -6 days before verifiable clinical signs or death were noticed, with an accuracy of 66-80% on day -3 and 80-91% on day -6. This indicates that continuous monitoring of animal locomotion is a superior predictor of animal health compared with human observation. The findings suggest that augmenting visual checks with AI modeling can greatly improve animal welfare by identifying subclinical cases, enhancing study endpoints, increasing the rigor and reproducibility of research, and improving operational efficiency. Our work underscores the potential of integrating advanced monitoring systems and AI in laboratory animal facilities, marking a substantial step forward in the field of animal welfare and research methodology.