Activity monitoring as a window into health and perfomance.

#monitoring #wind #embs

During the past 15 years, consumer wearable monitors have progressed rapidly from counting steps to a 
host of proprietary activity and sleep metrics. The next generation of monitoring capability will provide 
much more health-related, actionable information. In this talk, Brian will give an overview of these 
upcoming technologies and the underlying biosignal processing techniques for analyzing activity and 
sleep, including several examples from MIT Lincoln Laboratory: 1) early detection of Parkinson’s Disease 
and other movement disorders, 2) risk stratification based on levels of visceral adipose tissue, an 
important health biomarker, 3) presymptomatic detection of infectious disease, 4) quantitative 
estimation of fatigue to optimize scheduling of shift workers, 5) early detection of exertional heat stroke. 
The first two applications are based on analyzing publicly available accelerometry datasets: the U.S. 
National Health and Nutrition Examination Survey (NHANES) with about 8,000 subjects and the U.K. 
Biobank with 100,000 subjects. Biosignal processing algorithms that have been considered range from 
straightforward baselines of steps and cadence and sleep fragmentation, to correlation structure 
algorithms adopted from the radar signal processing community, to deep learning. All of these case 
studies involve close collaborations between engineers and senior physiologists, clinicians and end users; 
Brian will share experience on what is required for a high-performing engineering and medical team. 
Brian will also offer perspectives on wearable sensing technology transfer, progressing from idea to 
impact, and innovating in the national and global interest, based on his career working for a U.S. 
National Laboratory.