34
HEALTH AND INNOVATION I Jesus College Annual Report 2023
Dr Dimitris Spathis
Dr Ignacio Perez-Pozuelo
activities. By analysing data collected from wearable devices, such as heart rate monitors and accelerometers, researchers have now developed a way to accurately predict fitness levels. Researchers from the MRC Epidemiology Unit and the Department of Computer Science and Technology, worked across their disciplines to collect activity data from over 11,000 participants wearing the devices continuously over six days. The amount of data amassed was enormous – 60 data values per second was gathered from participants – requiring the design of an innovative algorithm pipeline to process the findings. Spathis said: “We had to design an algorithm pipeline and appropriate models that could compress this huge amount of data and use it to make an accurate prediction. The free-living nature of the data makes this prediction challenging because we’re trying to predict a high-level outcome (fitness) with noisy low-level data (wearable sensors).” The AI model they developed (known as a deep neural network) not only predicts VO2max from the gathered data, but also identifies sub-populations
needing specific fitness-related interventions. A machine learning model was developed and tested against a subset of these participants seven years later and was shown to produce more accurate data than previous methods. Co-lead author Perez-Pozuelo said: “This study is a perfect demonstration of how we can leverage expertise across epidemiology, public health, machine learning and signal processing”. Mascolo highlights the potential of their findings to redefine how we approach fitness measurement: “We’ve shown that you don’t need an expensive test in a lab to get a real measurement of fitness – the wearables we use every day can be just as powerful if they have the right algorithm behind them. Cardio-fitness is such an important health marker, but until now we did not have the means to measure it at scale.” The key to this innovation, according to Professor Mascolo, lies in transparency and the optimal use of data from heart rate and accelerometer sensors. She said: “This clarity, coupled with the ability to detect changes in fitness over time, enables the accurate estimation of fitness levels on a population-wide scale,