Google has developed an artificial intelligence model capable of predicting the risk of major adverse cardiovascular events (MACE) with similar success to hospital screenings, using just heart signals from photoplethysmographs (PPGs) or heart sensors commonly found in oxygen monitors or heart rate watches. This AI model is divided into two parts: the first being ResNet-18, which reads heart wave images from PPGs and predicts various patient characteristics without external input such as age, gender, body mass index, and high blood pressure symptoms. The data obtained is then combined with real patient information, such as gender, age, and smoking habits, to create a model that predicts the risk of encountering MACE within the next 10 years, relying on the UK Biobank dataset, which includes data from over 500,000 long-term patients, with approximately 200,000 having PPG data stored, sufficient for model creation.
The success of the prediction model is measured using the concordance index, demonstrating that the model can predict superiorly to using traditional risk prediction methods, eliminating the need for patients to undergo unnecessary screenings based on factors like blood pressure levels but instead using PPG sensors readily available for self-monitoring. The limitation of this research lies in the differences between the PPG data collected in the UK Biobank dataset and the sensors commonly used, prompting further study on the usability of alternative PPG sensors, such as mobile phone cameras, to potentially provide preliminary risk assessments to the general population.
TLDR: Google’s AI model accurately predicts the risk of major cardiovascular events using heart signals from photoplethysmographs, paving the way for personalized self-monitoring and early risk assessment.
Leave a Comment