Image

Cardiogram — Clinically Validated

Consumer wearables like Android Wear, Fitbit, and Apple Watch will generate two trillion health measurements this year—far too many for any human doctor to review. To help create the future of preventive medicine, we built DeepHeart, a novel deep neural network tested in multiple rigorous clinical studies. Below, you'll find four published, peer-reviewed research papers about DeepHeart. 

Your heart beats 102,000 times per day and reacts to everything that happens in your life—whether you're sick, your sleep quality, how you exercise, or your stress level.

DeepHeart: Semi-Supervised Deep Learning for Cardiovascular Risk Prediction

A major challenge for AI in medicine is that labeled training data is costly, scarce, and closely-guarded. DeepHeart is a semi-supervised deep neural network that accurately predicts cardiovascular risk, but requires 10x less labeled data than conventional deep learning techniques.

DeepHeart: Semi-Supervised Sequence Learning for Cardiovascular Risk Prediction. AAAI Conference on Artificial Intelligence, Feb 2018 (AAAI-18)

Image

Screening for Atrial Fibrillation

Atrial Fibrillation, the most common abnormal heart rhythm, causes 1 in 4 strokes and frequently goes undiagnosed. In the mRhythm Study, DeepHeart detected atrial fibrillation with 97% accuracy (c-statistic) using optical heart rate sensors, setting the stage for cost-effective, broadly-deployed AF screening.

Detecting Atrial Fibrillation using a Smart Watch - the mRhythm study Heart Rhythm Society Scientific Sessions, May 2017

Passive Detection of Atrial Fibrillation Using a Commercially Available Smartwatch JAMA Cardiology, March 2018

Image

Screening for Sleep Apnea and Hypertension

More than one billion people worldwide have hypertension and sleep apnea. In this study, DeepHeart detected both conditions with more than 82% accuracy using consumer wearables alone.

Cardiovascular Risk Stratification Using Off-the-Shelf Wearables and a Multi-Task Deep Learning Algorithm American Heart Association Scientific Sessions, Nov 2017.

Image

High Engagement Mobile Health Research

The first five ResearchKit apps lost 90% of their users in the first 90 days. By taking inspiration from popular consumer apps, we show it’s possible to improve retention by 5x, exceeding even Twitter and Instagram, with high retention across age groups.

Achieving High Retention in Mobile Health Research Using Design Principles Adopted From Widely Popular Consumer Mobile Apps. American Heart Association Scientific Sessions, Nov 2017

Image