Cambridge researchers have developed a method to measure overall fitness accurately on wearable devices — and more robust than current consumer smartwatches and fitness trackers — without the wearer needing to exercise.
Normally, tests to accurately measure VO2max – a key measure of overall fitness and an important predictor of heart disease and mortality risk – require expensive lab equipment and are mostly limited to elite athletes. The new method uses machine learning to predict VO2max – the body’s ability to perform aerobic work – during daily activity, without the need for contextual information such as GPS measurements.
In what is by far the largest study of its kind, researchers collected activity data from more than 11,000 Fenland study participants using wearable sensors, with a subset of participants tested again seven years later. The researchers used the data to develop a model to predict VO2max, which was then validated against a third group who performed a standard exercise test in the laboratory. The model showed a high degree of accuracy compared to laboratory tests and outperforms other approaches.
Some smartwatches and fitness trackers currently on the market claim to provide a VO2max estimate, but as the algorithms powering these predictions are unpublished and subject to change at any time, it is unclear if the predictions are accurate. or whether an exercise regimen has an effect on a person’s VO2max over time.
The model developed by Cambridge is robust, transparent and provides accurate predictions based solely on heart rate and accelerometer data. Since the model can also detect changes in fitness over time, it could also be useful for estimating fitness levels of entire populations and identifying the effects of lifestyle trends. The results are published in the journal npj Digital Medicine.
A VO2max measurement is considered the “gold standard” for fitness testing. Professional athletes, for example, test their VO2max by measuring their oxygen consumption as they train to exhaustion. There are other ways to measure fitness in the lab, such as heart rate response to stress tests, but these require equipment like a treadmill or exercise bike. Additionally, strenuous exercise can be a risk for some people.
VO2max isn’t the only measure of physical fitness, but it’s an important measure for endurance, and it’s a good predictor of diabetes, heart disease, and other mortality risks. However, since most VO2max tests are performed on reasonably fit people, it is difficult to obtain measurements from those who are not as fit and who may be at risk for cardiovascular disease.”
Dr Soren Brage, co-author, Cambridge’s MRC Epidemiology Unit
“We wanted to know if it was possible to accurately predict VO2max using data from a wearable device so that there would be no need for an exercise test,” said the co-lead author, Dr Dimitris Spathis from the Cambridge Department of Computing and Technology. “Our central question was whether wearables could measure fitness in nature. Most wearables provide metrics such as heart rate, steps, or sleep time, which are health indicators, but are not directly related to health outcomes.”
The study was a collaboration between the two departments: the MRC Epidemiology Unit team provided expertise in population health and cardiorespiratory fitness and data from the Fenland Study – a study of long-standing public health service in the east of England – while the team from the Department of Computing and Technology provided expertise in machine learning and artificial intelligence for mobile and wearable data.
Study participants wore wearables continuously for six days. The sensors collected 60 values per second, resulting in a huge amount of data before processing. “We had to design a pipeline of algorithms and proper models that could compress this massive amount of data and use it to make an accurate prediction,” Spathis said. “The free nature of the data makes this prediction difficult as we are trying to predict a high level outcome (fitness) with noisy low level data (wearable sensors).”
The researchers used an AI model known as a deep neural network to process and extract meaningful information from the raw sensor data and make VO2max predictions from it. Beyond predictions, trained models can be used for the identification of subpopulations in particular need of fitness-related intervention.
Baseline data from 11,059 Fenland study participants were compared with follow-up data seven years later, drawn from a subset of 2,675 original participants. A third group of 181 participants from the UK Biobank validation study underwent VO2max testing in the laboratory to validate the accuracy of the algorithm. The machine learning model showed strong agreement with VO2max scores measured at both baseline (82% agreement) and follow-up testing (72% agreement).
“This study is a perfect demonstration of how we can leverage expertise in epidemiology, public health, machine learning and signal processing,” said co-lead author Dr. Ignacio Perez-Pozuelo.
The researchers say their findings demonstrate how wearables can accurately measure fitness, but transparency needs to be improved if measurements from commercially available wearables are to be reliable.
“It’s true in principle that many fitness trackers and smartwatches provide a VO2max measurement, but it’s very difficult to assess the validity of these claims,” Brage said. “The models are generally not published and the algorithms can change regularly, making it difficult for people to determine whether their fitness has actually improved or is just being estimated by a different algorithm.”
“Anything on your smartwatch related to health and fitness is an estimate,” Spathis said. “We are transparent about our modeling and we have done it at scale. We show that we can achieve better results by combining noisy data and traditional biomarkers. In addition, all of our algorithms and models are open source and all the everyone can use them.”
“We’ve shown that you don’t need an expensive test in a lab to get a true measure of fitness – the wearables we use every day can be just as powerful, if they have the right algorithm behind them,” the lead author said. Professor Cecilia Mascolo from the Department of Computing and Technology. “Cardio fitness is such an important marker of health, but until now we didn’t have the means to measure it on a large scale. These results could have important implications for population health policies, so that we can move beyond weaker health indicators such as body mass index (BMI).”
The research was supported in part by Jesus College, Cambridge and the Engineering and Physical Sciences Research Council (EPSRC), part of UK Research and Innovation (UKRI). Cecilia Mascolo is a Fellow of Jesus College, Cambridge.
Spathis, D. et al. (2022) Longitudinal prediction of cardiorespiratory fitness using wearable devices in free-living environments. npj Digital Medicine. doi.org/10.1038/s41746-022-00719-1.
#method #accurately #measure #fitness #wearable #devices