Novel approaches to animal behaviour analysis using AI ～Advanced behavioural insights and precise quantification～

Challenge

Conventional animal behaviour analysis typically involves evaluators entering the animal housing room to directly observe animals or using specialized setups such as mazes or rotarod tests that differ from their normal housing environment.

While animal studies remain essential for new drug research and development, they present several challenges, including issues of reproducibility, objectivity, and their relevance to human biology.

Other limitations include the inability to monitor behavior during periods when staff are not present (such as at night) and the variability in assessment results that arises from differences in observer skill and experience.

LPIXEL’s Solution – AI Technology and Impact

By extracting anatomical key points from video—such as the ears, nose, limbs, and tail—a wide range of behavioural parameters can be derived, such as posture and movement speed. This enables both a comprehensive understanding of each specimen’s behavioral characteristics and precise quantitative analysis of specific actions (e.g., tremors or pruritic behavior).

For example, behaviors characteristic of disease model animals can be detected, allowing for more accurate and quantitative behavioral assessment in pharmacodynamic/pharmacological studies than is achievable with conventional methods.

Additionally, as part of the general observation of animals in safety testing, we aim to reliably identify drug-induced behavioral changes—such as altered activity levels or gait abnormalities.