Smartwatch Algorithm Detects Cardiac Arrest

The smartwatch algorithm, which leverages machine learning, was developed by Google Research (Mountain View, CA, USA), and is designed to meet performance standards suitable for large-scale use. The researchers first utilized photoplethysmography (PPG) to demonstrate that PPG measurements of peripheral pulselessness (induced by an arterial occlusion model) mirror the signal pattern of pulselessness seen in ventricular fibrillation (VF), a common arrhythmia associated with cardiac arrest. By analyzing the similarity between the PPG signals from VF and those from the occlusion model, they developed and validated a pulse loss detection algorithm using data collected from both the simulated model and real-world conditions.

Once the algorithm was developed, the researchers tested it in real-world scenarios. The results, published in Nature, showed that the system generated one unintentional emergency call per 21.67 user-years across two prospective studies. The algorithm’s sensitivity was 67.23% (with a 95% confidence interval of 64.32%–70.05%) when tested in a simulated arterial occlusion model of cardiac arrest. These findings indicate that this new technology offers a promising solution for wearable detection of sudden pulse loss while minimizing the societal impact of false alarms.