A new machine learning algorithm running on a smartwatch has shown the ability to detect sudden loss of pulse with remarkable accuracy, according to a study led by Google Research. The system, designed to recognize cardiac arrest, can automatically call for emergency help if the user becomes unresponsive.
The study found that the algorithm had a very high specificity (99.99%), meaning it rarely gave false alarms, and a moderate sensitivity (67.23%), indicating its ability to correctly identify cardiac arrest cases.
Cardiac arrest outside of a hospital, known as out-of-hospital cardiac arrest (OHCA), is a leading cause of sudden death. Immediate recognition and treatment are crucial for survival, but about 50–75% of these incidents happen when no one is around to help.
The researchers wanted to see if a smartwatch could detect when a person’s heart stops beating and automatically alert emergency services while keeping false alarms to a minimum.
The study, published in Nature, tested an algorithm that used data from a smartwatch’s photoplethysmography (PPG) sensor, which measures blood flow, along with motion sensors. To ensure accuracy, the researchers tested the system in different environments, including medical settings and real-world conditions.
In a controlled medical lab, 100 patients undergoing heart procedures had their heart rhythm deliberately stopped for a short time, allowing researchers to collect real data on pulselessness.
In another test, 99 people had their blood flow temporarily cut off using a tourniquet to simulate the loss of a pulse. Additionally, 948 participants were monitored in their daily lives to measure the smartwatch’s ability to avoid false alarms.
To further test the algorithm’s reliability, 220 participants wore the smartwatch during their normal routines, while 135 of them also participated in controlled experiments where their pulse was intentionally stopped.
Another 21 professional stunt performers simulated real-life collapses caused by cardiac arrest to see how the algorithm responded to sudden motion events.
The results showed that the smartwatch was highly accurate in recognizing when a person lost their pulse. The system detected pulselessness within 57 seconds and waited an additional 20 seconds for a user response before placing an emergency call.
The algorithm’s sensitivity varied depending on the situation—72% for cases where the person was motionless and 53% for cases where the person collapsed suddenly. However, it had an exceptionally low false alarm rate, with only one mistaken emergency call expected per 21.67 years of smartwatch use.
Smartwatches with this capability could significantly improve survival rates for people who experience cardiac arrest when they are alone. Early detection and quick medical response are key factors in saving lives. However, to ensure that emergency calls are only made when necessary, researchers are working on reducing the number of false alarms.
Since the algorithm was trained on controlled events, it may not perfectly detect every real-life case of cardiac arrest. Ongoing data collection from smartwatch users in real-world settings will help improve the system’s accuracy and reliability in different conditions.
With further development, this technology could become an essential tool for detecting life-threatening emergencies and providing faster medical assistance to those in need.
If you care about heart health, please read studies about how vitamin D influences cholesterol levels, and what we know about egg intake and heart disease.
For more health information, please see recent studies about best supplements for heart disease prevention, and wild blueberries can benefit your heart and brain.
The research findings can be found in Nature.
Copyright © 2025 Knowridge Science Report. All rights reserved.
