California-based startup Cardiogram is using data from popular fitness trackers and smartwatches to detect severe health conditions.
Around 20% of adults own some sort of wearable electronic device that monitors health. Smartwatches like the Apple Watch Series 6 and fitness trackers such as the Fitbit collect an immense amount of data about the wearer’s vitals and physical activity. Oxygen saturation, heart rate, sleep quality, and exercise are just a few of the data points that some wearables continuously record throughout the day and night.
Smartwatches and fitness trackers are accompanied by mobile apps and online dashboards that smartly display the data in easy-to-understand graphs. The information is used to help the wearer track their exercise progression and look for ways to improve their overall wellness. However, what if that same data could be used to detect serious health conditions?
Cardiogram is an app for iOS and Android that does just that. It collects data from a paired Garmin, FitBit, WearOS, or Apple device to look for data anomalies that require the attention of a medical professional.
The researchers at Cardiogram used the data of over 9000 study participants to develop and train their deep neural network, DeepHeart. DeepHeart looks for several indicators in the user’s data that there may be an underlying issue: heart rate spikes at rest, heart rate variability, heart rate recovery, and more. For example, a study in 2017 found that DeepHeart could detect atrial fibrillation with 97% accuracy. The app can also detect hypertension or high blood pressure, sleep apnoea, and diabetes.
Wearables don’t always display data efficiently enough to be useful to healthcare providers, and Cardiogram has kept that in mind. For example, whenever there is an unusual spike in heart rate at rest, the app tags the spike and allows the user to find that information quickly. Likewise, users can mark periods in their heart rate with notes to record symptoms associated with heart rate changes. The app automatically keeps anomalies to be easily searchable and shareable with the user’s doctor.