Watch Breathing Rate Lacks Validation During Exercise
A chest strain sensor stayed within 3 breaths a minute in 24 healthy people walking. No consumer watch was tested in these studies.
Evidence rating for the proposition “Consumer-grade wearable devices measure respiratory rate with accuracy comparable to clinical reference methods”: Uncertain
Not established for consumer watches. The studies found here tested research sensors in small groups of healthy volunteers. A stretchable strain sensor on the ribcage and abdomen stayed within 3 breaths per minute of a flow sensor in 24 people during a walking test. A face-mask microphone in 10 people had error up to 3.8 breaths per minute while running at 12 km/h. Pulse-based algorithms had limits of agreement of -5.1 to 7.2 breaths per minute under ideal conditions.
Do chest sensors keep up during exercise?
In a walking test, yes. Twenty-four healthy volunteers wore stretchable strain sensors on the ribcage and abdomen. The test included standing, sitting, walking and walking with a stick. A standard flow sensor was the reference.
The limits of agreement stayed within 3 breaths per minute throughout. The agreement held regardless of breathing rate or volume. The authors say it produced accurate breath counts during moderate exercise in healthy young people.
Sensor type matters when the body moves. A gym study compared four indirect wearable systems against a spirometer during four exercises. All were affected by movement artifacts. Piezoelectric pneumography was the most robust in three of the four exercises.
2 studies
- In 24 healthy volunteers wearing stretchable strain sensors on the ribcage and abdomen during a walking protocol, limits of agreement against a flow sensor stayed within 3 breaths per minute, regardless of respiratory rate or volume.
- In a gym study of four indirect wearable breathing systems against a spirometer, all were affected by movement artifacts, and piezoelectric pneumography was most robust in three of four exercises.
What about microphones and motion sensors?
A face-mask microphone was tested in 10 healthy people. They rested, walked at 3 and 6 km/h and ran at 9 and 12 km/h. An orifice flowmeter was the reference. Error rose with exercise intensity and noise, up to a mean absolute error of 3.8 breaths per minute while running at 12 km/h. Across all conditions the mean absolute error was 1.7 and the limits of agreement were -0.24 plus or minus 5.07.
A three-sensor motion system on the thorax, abdomen and lower back was tested in 20 healthy people. It detected significant differences in breathing rate between static postures and slow walking, fast walking, running and cycling, in line with physiology. The abstract reports no error figure against a reference.
Are chest-worn devices consistent across repeat sessions?
Yes, in one large healthy sample, but that is consistency and not accuracy. Sixty healthy adults wore the Zephyr BioHarness at rest, during a submaximal aerobic fitness test and in recovery. Test-retest reliability was 0.76 to 0.84 at rest, 0.76 to 0.89 during the test and 0.79 to 0.90 in recovery.
The authors call the reliability excellent. A device can be consistent and still be wrong. The study did not test maximal exercise.
What about breathing rate estimated from a pulse or ECG signal?
Only under ideal conditions. A 2016 study tested 314 algorithms on healthy participants against a nasal-oral pressure sensor. The best ECG-based algorithm had 95% limits of agreement of -4.7 to 4.7 breaths per minute. The best pulse-based one had -5.1 to 7.2 and a bias of 1.0.
Algorithms did better with ECG than with the pulse signal. The abstract reports ideal conditions and does not report exercise results. A review of these algorithms frames them as an opportunity for healthcare and fitness monitoring and calls for research on their use in wearable sensors.
For the resting picture, see whether wearables measure respiratory rate accurately. For placement, see wrist versus chest placement.
What does this not settle?
No source here tests a consumer watch, ring or strap for breathing rate during exercise against a reference. The exercise studies used research prototypes and 10 to 60 healthy volunteers.
Nothing here covers maximal effort, sports with heavy arm movement or people with lung disease.
The exercise results come from research sensors tested in 10 to 60 healthy volunteers at walking, running and submaximal effort. None validates a named consumer watch or ring, and none tests maximal exercise or people with lung disease.
Common questions
How accurate is wearable breathing rate during exercise?
A research strain sensor stayed within 3 breaths per minute of a flow sensor in 24 healthy volunteers during a walking test. A face-mask microphone had error up to 3.8 breaths per minute at 12 km/h running in 10 people.
Has a consumer watch been tested for breathing rate during exercise?
Not in the sources reviewed here. They tested research sensors in small groups of healthy volunteers.
Is a pulse-based breathing rate accurate?
Under ideal conditions in healthy people, the best pulse-based algorithm had limits of agreement of -5.1 to 7.2 breaths per minute, wider than the best ECG-based one at -4.7 to 4.7.
What is the difference between reliability and accuracy?
Reliability is repeatability across sessions. The Zephyr BioHarness gave consistent breathing rate readings in 60 adults, but a consistent device can still read wrong against a reference.
Does movement add error to breathing sensors?
Yes. In a gym study, all four wearable breathing systems were affected by movement artifacts, though piezoelectric pneumography held up best in three of four exercises.
Sources
- Monitoring respiratory rates with a wearable system using a stretchable strain sensor during moderate exercise.
- Comparative evaluation of susceptibility to motion artifact in different wearable systems for monitoring respiratory rate.
- Respiratory Rate Estimation during Walking and Running Using Breathing Sounds Recorded with a Microphone.
- An IMU-Based Wearable System for Respiratory Rate Estimation in Static and Dynamic Conditions. [conflict statement: e-Novia part-funded the work and owns a patent application naming two authors]
- Reliability of Zephyr BioHarness Respiratory Rate at Rest, During the Modified Canadian Aerobic Fitness Test and Recovery.
- An assessment of algorithms to estimate respiratory rate from the electrocardiogram and photoplethysmogram.
- Breathing Rate Estimation From the Electrocardiogram and Photoplethysmogram: A Review.