DeepHeart: Semi-Supervised Sequence Learning for Cardiovascular Risk Prediction

TL;DR

This paper introduces a semi-supervised LSTM model trained on wearable sensor data to accurately predict multiple cardiovascular-related health conditions, outperforming traditional biomarkers and enabling new risk stratification methods.

Contribution

The paper presents a novel semi-supervised multi-task LSTM approach for health condition detection using wearable data, surpassing existing biomarker methods.

Findings

Achieved high accuracy in detecting diabetes, high cholesterol, high blood pressure, and sleep apnea.

Semi-supervised training methods outperform hand-engineered biomarkers.

Demonstrates potential for risk stratification using consumer wearable devices.

Abstract

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