Abstract

We train and validate a semi-supervised, multi-task LSTM on 57,675person-weeks of data from off-the-shelf wearable heart rate sensors, showinghigh accuracy at detecting multiple medical conditions, including diabetes(0.8451), high cholesterol (0.7441), high blood pressure (0.8086), and sleepapnea (0.8298). We compare two semi-supervised train- ing methods,semi-supervised sequence learning and heuristic pretraining, and show theyoutperform hand-engineered biomarkers from the medical literature. We believeour work suggests a new approach to patient risk stratification based oncardiovascular risk scores derived from popular wearables such as Fitbit, AppleWatch, or Android Wear.

 

Quick Read (beta)