Cardiogram, a startup working on algorithms to make the Apple Watch’s heart rate data clinically actionable, announced some results today from its mRhythm Study. The data, presented at the Heart Rhythm Society’s 38th Annual Scientific Sessions, shows that the company’s algorithms can detect atrial fibrillation with 97 percent accuracy.
“Our results show that common wearable trackers like smartwatches present a novel opportunity to monitor, capture and prompt medical therapy for atrial fibrillation without any active effort from patients,” Senior Author Dr. Gregory M. Marcus, endowed professor of atrial fibrillation research and director of clinical research for the division of cardiology at UCSF, said in a statement. “While mobile technology screening won’t replace more conventional monitoring methods, it has the potential to successfully screen those at an increased risk and lower the number of undiagnosed cases of AF.”
Atrial fibrillation (AF) is a common heart arrhythmia that affects more than 2.7 million Americans. Detecting it has been the focus of health startups like Cardiogram and smartphone ECG company AliveCor because AF can lead to a stroke, but is often asymptomatic.
To create the algorithm, Cardiogram, in coordination with UCSF’s Health eHeart Study, enrolled 6,158 users of the Cardiogram Apple Watch app. Data including heart rate and mobile ECGs was collected from those patients and used to train a deep neural network.
A cohort of 51 patients scheduled for a procedure called cardioversion, which restores heart rhythms after an arrythmia, used both an Apple Watch and a 12-lead ECG before and after the procedure. The neural network-derived algorithm identified AF with 97 percent accuracy, 98 percent sensitivity, and 90.2 percent specificity.
Researchers will continue to hone the algorithm and also explore the potential of using a deep neural network to detect other conditions.
Cardiogram raised $2 million last October in a round led by Andressen Horowitz's new bio fund, led by Vijay Pande.