Key result
DeepHeart, a semi-supervised multi-task LSTM trained on wearable heart rate data, achieved high accuracy in detecting diabetes (AUC 0.8451), sleep apnea (0.8298), hypertension (0.8086), and high cholesterol (0.7441).
Why the study?
Does a semi-supervised, multi-task LSTM using wearable heart rate sensor data improve the detection of cardiovascular risk factors compared to hand-engineered biomarkers?
Observational (n=14,011)
Does a semi-supervised, multi-task LSTM using wearable heart rate sensor data improve the detection of cardiovascular risk factors compared to hand-engineered biomarkers?
Deep learning applied to consumer wearable heart rate data can accurately detect multiple cardiovascular risk factors, offering a novel approach to patient risk stratification.
May support passive wearable screening for cardiometabolic risks; hypothesis-generating and requires prospective validation before clinical adoption.
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 training 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.
No takes yet. Share an insight, caveat, or question.
Ballinger et al. (2018) conducted an observational in Cardiovascular risk factors (diabetes, hypertension, sleep apnea, high cholesterol) (n=14,011). DeepHeart (Semi-supervised multi-task LSTM) vs. Hand-engineered biomarkers and standard machine learning algorithms was evaluated on Accuracy (c-statistic/AUC) for detecting diabetes, high cholesterol, high blood pressure, and sleep apnea. DeepHeart, a semi-supervised multi-task LSTM trained on wearable heart rate data, achieved high accuracy in detecting diabetes (AUC 0.8451), sleep apnea (0.8298), hypertension (0.8086), and high cholesterol (0.7441).
Synapse has enriched 3 closely related papers on similar clinical questions. Consider them for comparative context: