Comparing Radar and Force-Sensor Method agreement for Non-Invasive Respiratory Monitoring of Freely-Moving Mice

Abstract

Continuous non-invasive monitoring of respiratory rate in laboratory mice can support research across drug efficacy, sleep, and stress and pain assessment. Established respiratory rate estimation methods exist but many conflict with the 3Rs principle through invasiveness, motion restriction, or animal stress. A cross-method comparison between two non-invasive systems on the same freely-moving animal offers a verification path that does not require an invasive reference. The goal was to conduct such a comparison, analyzing agreement between methods and when methods become unreliable dependent on behavioral state.

A signal-processing pipeline was developed for a 60 GHz pulse coherent radar
(Acconeer A121) mounted above a home cage, extracting respiratory rate from
chest-wall-displacement phase information. Twelve pipeline iterations were each accepted against a non-regression gate on a synthetic-signal benchmark.

Resulting respiration rate estimates from the final pipeline were compared against an existing force-sensor respiratory rate metric (TrackPaw) on simultaneous recordings spanning four cages and multiple 24-hour long sessions, with neither system treated as ground truth. Agreement was characterized by Bland-Altman analysis with a repeated-measures formulation. Behavioral-state robustness was characterized by per-method coverage and by agreement between the radar flagging motion and TrackPaw's flagging travel.

After 30 s pre-pairing aggregation, the joint bias between the methods was +0.03 Hz with 95 % limits of agreement of −0.14 and +0.20 Hz. Neither method reliably produced estimates during active behavior, when radar and TrackPaw both flag motion, and most coverage from both methods happened when no motion was flagged.

Continuous non-invasive respiratory rate monitoring in the home cage appears feasible at rest. Reliable estimation during active behavior remains an open problem for both methods.

Popular Abstract

Watching mice breathe using floors and invisible energy

How can a mouse's breathing be measured without touching, holding, or even watching the animal? Two very different sensors did exactly that, in parallel, and agreed on the answer.

Breathing tells researchers a lot. A drug that quiets pain may slow breathing. A poor night's sleep can speed it up. In studies of asthma, anxiety, sleep, or heart disease, the breathing rate of a laboratory mouse can be one of the most informative numbers in the experiment. The trouble is collecting that number without disturbing the animal's regular behavior, impacting what the measurement really tells us.

Each established method comes with a price. Sealed chambers force the animal to sit still and can stress the mouse into changing the very breathing being measured. Implanted sensors require surgery and can leave lasting effects on the animal. Even watching with a stopwatch in hand burns human hours and adds an observer that mice can sense and react to.

This thesis tested two methods that quietly read breathing while the mouse roams freely in a home cage. The first is a small radar mounted above the cage that sends short pulses of high-frequency radio waves through the lid and listens for the faint echo from the chest wall rising and falling. The second is a platform of force sensors under the cage floor that picks up the tiny force shifts between an in-breath and an out-breath pushing and pulling on the floor. Neither sensor ever touches the animal, and the mouse is free to walk, climb, eat, and sleep.

There is a twist. Neither device can be called the truth. So the question changes from whether one device gets the right answer to whether two sensors, built on completely different physical principles and knowing nothing about each other, will agree. Such an agreement would be hard to explain unless both happen to be reading the same underlying signal.

Across simultaneous recordings from four cages and many hours of mouse rest, sleep, and ordinary activity, the two methods landed within roughly 10 breaths per minute of each other when measuring once every 30 seconds. That might sound like a lot until you realize mice typically breathe near 200 breaths per minute! That is a tight match for entirely independent technologies. When the mouse moved, however, both methods went blind in similar ways. The radar lost the breathing rhythm in the clutter of small body movements, and the force-sensor platform could not separate breathing when a mouse was running around.

Quietly counting breaths around the clock may help squeeze more usable data out of every animal study, lowering the number of mice needed to answer the same scientific question. That fits the principle of 3Rs, the call to replace, reduce, and refine the use of laboratory animals. The same kind of unobtrusive monitoring may, in time, support long-term welfare checks in many other animal facilities. The mouse, meanwhile, never knows the device is there.

• author: Andersson, Axel
• supervisor: Frida Sandberg; Erik Månsson
• organization: Division for Biomedical Engineering
• alternative title: Jämför överensstämmelse av radar- och kraftsensorsystem för ickeinvasiv mätning av andning på frigående möss
• course: BMEM01 20261
• year: 2026
• type: H2 - Master's Degree (Two Years)
• subject: Technology and Engineering
• language: English
• additional info: 2026-11
• id: 9238185
• date added to LUP: 2026-06-29 09:12:19
• date last changed: 2026-06-29 09:12:19