Wearables, smartphones, and artificial intelligence for digital phenotyping and health
Ubiquitous progress in wearable sensing and mobile computing technologies, alongside growing diversity in sensor modalities, has created new pathways for the collection of health and well-being data outside of laboratory settings, in a longitudinal fashion. Wearable and mobile devices have the potential to provide low-cost, objective measures of physical activity, clinically relevant data for patient assessment, and scalable behavior monitoring in large populations. These data can be used in both interventional and observational studies to derive insights regarding the links between behavior, health. and disease, as well as to advance the personalization and effectiveness of commercial wellness applications. Today, over 400,000 participants have had their behavior tracked prospectively using accelerometers for epidemiological studies across the globe. Traditionally, epidemiologists and clinicians have relied upon self-report measures of physical activity and sleep which, while valuable in the absence of alternatives, are subject to bias and often provide partial, incomplete information Physical behavior data extracted from wearable devices are being used to derive sensor-assessed, objective measures of physical behaviors, overcoming the limitations of self-report with the aim of relating these to clinical endpoints and eventually applying the findings to preventive and predictive medicine. Moreover, the application of artificial intelligence (AI), sensor fusion, and signal processing to wearable sensor data has led to improved human activity recognition and behavioral phenotyping. Here, we review the state of the art in wearable and mobile sensing technology in epidemiology and clinical medicine and discuss how AI is changing the field.