Apple Watch collects a tremendous amount of health data and apple, with researchers, still finding out what this data can provide features. Now, it seems that the next Apple Watch can see the pregnancy test as a new Apple Watch feature because researchers have found that watch -collecting data can be used with some great accuracy for this purpose.

According to a new published study, an AI model developed with support from Apple is capable of detecting health conditions such as pregnancy with remarkable accuracy, nothing more than behavior data collected from equipment such as Apple Watch and iPhone.

Studies, Beyond Sensor Data: Foundation models of behavior data from wearbals improve health predictions – and reported by 9to5MAC – introduces a new type of machine learning model known as wearable behavior models (WBM). Unlike traditional health models, which rely on raw sensor data, such as heart rate or blood oxygen levels, WBM are allegedly trained on long -term behavior patterns. These include activity level, sleep quality, mobility, heart rate variability and other high-level health matrix that have already been calculated by the Apple Watch algorithm.

The WBM was developed as part of the Apple Heart and Movement Study (AHMS), which included over 160,000 participants who voluntarily shared their health data. Overall, the researchers trained the AI ​​model on a huge dataset, which included more than 2.5 billion hours of wearable data. The model was tested in 57 different health prediction tasks and according to researchers, AI continuously performed better than traditional sensor-based models-especially when identifying subtle changes in the body associated with conditions such as pregnancy, infection, or recovery of injury.

How does this work?

WBM allegedly focuses on users’ behavior rather than short -term biometric spikes. The model has analyzed the week-by-week patterns using the next generation architecture, called the Mamba-2, which is particularly effective for human routine such as time-series data. This enables it to detect subtle, cumulative changes that can indicate anything from poor sleep to respiratory infections or even early stage pregnancy.

AI model used Apple Watch to detect pregnancy

WBM, especially to detect pregnancy, obtained 92 per cen accurate with daily insight and traditional biometric data such as PPG (photoplathysmography). This hybrid has proved the approach particularly effective, crossing the model that completely depends on the raw heart rate or oxygen levels. In fact, researchers found that behavioral matrix such as changes in gate, mobility pattern, and sleep duration were among the most reliable indicators of early pregnancy.

In particular, the study does not propose to change the raw sensor data completely. Rather, it advocates a joint approach. While real -time biometric data can detect acute physical phenomena, behavior data said to provide a more comprehensive picture of long -term health trends. This dual-input method was also found to be the most effective in cases ranging from detection of respiratory infection, whether any specific drugs are on, such as beta blockers.