Article révisé par les pairs
Résumé : The quantified-self of a person consists in the self-tracking of health and physiological parameters, such as (but not limited to) heart rhythm, energy expenditure, and sleep, using technology and devices such as smartwatches or wristbands, without the need of being supervised by clinicians. The widespread adoption in recent years of wearables, combined with the increased relevance of Internet of Things in healthcare and in exercise equipment, have made self-tracking accessible to a large segment of the population with various performances. In this study, we present the development, optimization, and preliminary validation of a new device aimed to analyze sleep, activity level, and energy expenditure. Based on the results measured with a sensor previously certified (but now out-of-the-market), a machine learning model was trained and validated showing a very satisfying agreement of the results. In order to optimize the software and check the measurement accuracy in vivo, a clinical study on 12 healthy volunteers was performed comparing the results measured by the device with the one obtained by a metabolimeter considered as the gold-standard. The results demonstrated that the device is able to correctly assess energy expenditure, showing a difference lower than 19% of the value given by the gold standard.