Contactless metabolism estimation of small animals using high-frequency millimeter-wave radar

Hiroaki Ono,Kiyomi Ishikawa,Ayaka Wataki,Shoko Fujino,Genshiro A. Sunagawa

DOI: https://doi.org/10.1101/2024.04.11.588816

2024-04-14

Abstract:Animals flexibly adapt to internal and external environmental changes by utilizing energy produced from oxygen as fuel. By non-invasively monitoring an animal’s oxygen consumption, it becomes possible to understand an individual’s metabolic state. Calorimeters are known for directly measuring oxygen consumption but come with the issue of high initial costs. Despite the development of non-invasive techniques for measuring vital signs—including respiration rate (RR), heart rate (HR), and body temperature (Tb)—as indicators of metabolism, conventional methods encounter difficulties in estimating oxygen consumption rate (VO2). In this study, we developed a system that estimates the oxygen consumption of small animals using signals obtained from millimeter wave (mm-wave) radar technology processed through machine learning. By identifying frequency bands within the mm-wave signals contributing to VO2, our system is capable of predicting oxygen consumption several minutes in advance. Our system enables contactless, low-cost and multiplexed measurements of oxygen consumption, presenting a significant advancement in the field.

Physiology

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to non - contactedly estimate the metabolic rate (especially oxygen consumption \(VO_2\)) of small animals by using millimeter - wave radar technology. Traditional metabolic rate measurement methods such as direct or indirect calorimetry are accurate but costly and have a slow response time, while other non - contact monitoring techniques (such as video recording technology) have limitations in accuracy, especially when the animals are occluded. In addition, the existing millimeter - wave radar technology can effectively extract vital sign signals (such as respiration rate and heart rate), but has not been successfully used to estimate the metabolic rate.

To solve these problems, the researchers developed a machine - learning - based algorithm, which uses the features extracted from millimeter - wave radar signals to predict the immediate \(VO_2\) of small animals and further predict the future \(VO_2\). This technology can not only achieve non - contact, low - cost and multiplexed \(VO_2\) measurement, but also predict oxygen consumption several minutes in advance, which is of great significance for studying metabolic dynamics and adaptability.