Apple-Backed AI Model Achieves 92% Accuracy in Pregnancy Detection Using Wearable Data

A new artificial intelligence model, developed with support from Apple and trained on 2.5 billion hours of wearable data from Apple Watch and iPhone, demonstrates significant advancements in health prediction. While a recent social media post by Sar Haribhakti highlighted the model's ability to flag "heart conditions with 92% accuracy," detailed research indicates the 92% accuracy figure specifically pertains to pregnancy detection when combined with traditional biometric data. The model also shows improved detection capabilities across a range of other health conditions.

This innovative system, known as the Wearable Behavior Model (WBM), was developed as part of the Apple Heart and Movement Study (AHMS). Researchers trained WBM on an immense dataset comprising behavioral metrics from over 160,000 participants. Unlike traditional models that primarily rely on raw sensor data like heart rate or blood oxygen, WBM analyzes higher-level behavioral patterns such as step count, sleep duration, and mobility.

The study, titled "Beyond Sensor Data: Foundation Models of Behavioral Data from Wearables Improve Health Predictions," revealed WBM's strong performance across 57 health-related tasks. In addition to its high accuracy for pregnancy detection, the model demonstrated consistent improvements in identifying static health states like medication use and dynamic conditions such as sleep quality, respiratory infections, and cardiovascular issues like Afib detection. This approach complements, rather than replaces, raw sensor data by focusing on long-term behavioral trends.

The integration of this AI model could significantly enhance the health monitoring capabilities of Apple's wearable devices. Researchers note that existing Apple Watch hardware is capable of supporting this advanced AI-powered analysis. The model's ability to interpret subtle changes in user behavior over time presents a promising avenue for earlier detection and management of various health conditions, potentially transforming how individuals manage their well-being through everyday technology.