We will discuss about \u201cWearable data science for personalized digital medicine\u201d<\/p>\n
Abstract<\/p>\n
Millions of people currently use wearables such as the Apple Watch to monitor their physical activity, heart rate, and other physiological signals, generating an unprecedented amount of wearable data. This presents an opportunity for digital medicine to advance precision medicine. However, the noisy nature of this wearable data makes it appear unusable without new mathematical techniques to extract key signals from it. In this talk, I will discuss several techniques we have developed for analyzing this noisy time-series data, including the level-set Kalman filter-based data assimilation technique – a new state space estimation method that can estimate the phase of circadian rhythms. Additionally, I will introduce a Kalman filter-assisted autoencoder used for anomaly detection in time-series data, as well as feature engineering based on persistent homology and mathematical modeling. These techniques have practical applications, such as sleep scoring, detection of physiological changes related to COVID-19, and daily mood prediction.<\/p>\n","protected":false},"excerpt":{"rendered":"
We will discuss about \u201cWearable data science for personalized digital medicine\u201d Abstract Millions of people currently use wearables such as the Apple Watch to monitor their physical activity, heart rate, … <\/p>\n