Using Machine Learning and Predictive Artificial Intelligence to Determine Cage Change Frequency for Mice Housed in...

Area of Science:

• Laboratory Animal Science
• Animal Welfare Technology
• Artificial Intelligence in Research

Background:

• Standard mouse cage change frequency (2-wk) relies on subjective visual assessment, often necessitating premature changes.
• Current methods lack objective metrics, leading to potential inefficiencies and resource waste in research facilities.

Purpose of the Study:

• To validate a digital monitoring technology for objectively determining mouse cage change necessity.
• To assess the impact of an extended cage change schedule on animal health and environmental parameters.

Main Methods:

• Development and training of a machine learning/artificial intelligence algorithm correlating human observations with a digital Bedding Status Index (BSI).
• Validation of the algorithm across diverse mouse strains, ages, sexes, and cage densities.
• Comparison of extended cage change intervals with standard practices and assessment of key welfare and environmental indicators.

Main Results:

• The digital monitoring system achieved >90% accuracy in identifying soiled cages for higher densities (5 animals/cage).
• Average cage change intervals extended to 3-6 weeks, significantly longer than the standard 2 weeks, with reduced accuracy for single-housed mice (76%).
• Extended intervals did not negatively affect intracage ammonia/CO2 levels, mouse growth rates, or circadian rhythms, leading to a 65-70% reduction in cage changes.

Conclusions:

• The Bedding Status Index (BSI) provides an objective marker for mouse cage changes, enabling data-driven decisions.
• Digital monitoring technology significantly enhances operational efficiency by reducing cage changes, labor, and resource consumption.
• This technology supports improved animal welfare through optimized environmental conditions and reduced handling stress.