"AI Eye" Monitors Animal Experiments 24/7! NHRI's Latest Invention Advances More Humane Pharmaceutical Research

The National Health Research Institutes (NHRI) has partnered with National Tsing Hua University to develop an 'AI image recognition system for precise identification of experimental animal behaviors.' Utilizing video recording, deep learning, and AI recognition technologies, this system enables continuous, real-time observation of laboratory animal activities. It can identify at least eight distinct behaviors—such as drinking, grooming, eating, searching, resting/sleeping, gnawing, scratching, and rearing—24 hours a day, allowing researchers to not only 'see' but also 'measure' animal behavior with data tracking and analysis capabilities. This research has been published in the internationally renowned scientific journal iScience.

Biomedical research frequently relies on validation across cell experiments, animal testing, and human clinical trials. Physiological and behavioral changes in experimental animals serve as critical indicators for evaluating the safety and efficacy of treatments. However, traditional behavioral observation methods require extensive manual recording by researchers, which is time-consuming and prone to human error and observational bias, especially under long-term, continuous monitoring demands.

To address this, NHRI’s Institute of Biomedical Engineering and Nanomedicine, in collaboration with its Animal Research Center and National Tsing Hua University, has integrated cross-institutional expertise to transform animal behaviors into trackable, analyzable data using AI technology. The team has developed a multi-functional imaging recorder that can be installed in standard animal housing cages. Cameras continuously record animal activities, while behavior classification models powered by deep learning and AI algorithms analyze the footage.

Currently, the system can continuously identify at least eight types of behaviors over 24 hours, with recognition accuracy exceeding 90% for four behaviors and over 80% overall. This advancement supports the internationally recognized '3Rs principles'—Reduction, Replacement, and Refinement—which form the ethical foundation of animal research. By enabling real-time, quantitative monitoring of animal behavior, the system helps researchers continuously assess animal well-being and establish measurable quality-of-life indicators, advancing the refinement of animal testing practices.

Integrated with wireless network infrastructure and robust information security protocols, the system enhances data integrity and the reliability of biomedical research outcomes while significantly reducing the manpower required for prolonged manual observation. It provides a new technological tool for improving both animal care and research management.

The research team emphasizes that this breakthrough demonstrates the vast potential of AI in biomedical research. With broader interdisciplinary collaboration, talent development, and continuous training, AI applications in life sciences can expand further, accelerating the translation of research into practical outcomes. NHRI has been actively promoting interdisciplinary research and cultivating talent through institutes such as the National Institute of Cancer Research, the National Institute of Environmental Health, and the Institute of Biomedical Engineering and Nanomedicine, offering integrated programs in engineering, science, and molecular imaging.

Moving forward, NHRI aims to strengthen cross-domain partnerships, ensuring that AI not only serves as a key technical tool in pharmaceutical R&D but also enhances the refinement, ethical standards, and reliability of animal experimentation—opening new possibilities for biomedical innovation.