The Apple Watch could be used to detect a heart condition that causes over 100,000 strokes every year, according to a new study.
Heart health app Cardiogram and researchers from the University of California, San Francisco (UCSF) Cardiology Health eHeart project teamed up to take a closer look at just how effective the Watch can be at tracking the most clinically common heart abnormality, atrial fibrillation (AF).
The irregularity, which is treatable but tough to diagnose using current medical standard practices, is the leading cause of heart failure.
The mRhythm project that resulted from the pairing looked at the Apple Watch-sourced heart rate readings from 6,158 Cardiogram users.
The data was then used to build an algorithm to detect the distinct heart rate variability pattern caused by AF.
The team used an AI technique called semi-supervised deep learning to train a neural network to sift through the data to identify the heart rate irregularities.
The method is similar to the recent work at Stanford that used an AI neural network to identify skin cancer, although that study depended on one of Google's image recognition algorithms, not an entirely new one.
After being trained, the research team's algorithm was able to detect atrial fibrillation accurately 97 per cent of the time, which Cardiogram software engineer Avesh Singh claims beat existing methods of diagnosis in a blog post.
The study's results were presented for the first time at the 2017 Heart Rhythm conference.
"The most promising finding of our study is proof that consumer-grade wearables can be used to detect disease," Singh wrote.
"The future is bright here, and there are a few research directions that are particularly interesting to us."
Read the full article here.