Researchers from the University of Cambridge along with the Alan Turing Institute have published a study in which they have shown that an Artificial Intelligence (AI) model trained on wrist accelerometer and wearable electrocardiogram (ECG) data can successfully predict a person’s physical profile including physical fitness and health characteristics.
Termed the “Step2Heart”, a general-purpose self-supervised feature extractor for wearable data, which leverages the multimodal nature of modern wearable devices to generate participant-specific representations
The researchers state that the current wearable devices only record primitive information like step count and heart rate using low-level signals like acceleration. Leveraging this access to information, they can now use this information to profile a person using the data from their devices instead of going the other way around.
Their model is based on data from 2,100 individuals between the age of 35 and 65 comprising of more than 280,000 hours of wrist accelerometer & wearable ECG data.
This could be the next step for healthcare as it integrates existing technology to profile humans to the extent where it can predict basic health-related outcomes. It only has a 70% AUC which is the ratio of true positives to false positives.
Image Source: Information Age
