30 minutes into the future: AI warns of cardiac arrhythmia before it occurs

If the heart's movements get out of sync, it can quickly become threatening. From an impending heart attack to dementia, numerous diseases are directly linked to cardiac arrhythmias such as atrial fibrillation.

So it's no wonder that smartwatches from Apple, Fitbit and Samsung can already detect such deviations with almost 100% accuracy with the support of machine learning.

However, a study by the University of Luxembourg has now gone one step further. Instead of training an AI with the measurements of an electrocardiogram (ECG) over a short period of time, significantly longer periods of time should be recorded.

These long-term measurement series can be realized much more easily as pure heart rate measurements, for example. Current smartwatches and fitness wristbands can also record this value very precisely. The ECGs that smartwatches also record, on the other hand, are only reliable to a limited extent.

With 24-hour data sets from several hundred patients, the researchers were then able to predict cardiac arrhythmias an average of 30 minutes before they occur.

The reliability with pure heart rate measurement was even slightly better than with an ECG. In each case, AI was able to warn of atrial fibrillation and other disorders with more than 80 percent accuracy.

Without individual outliers in the data, it was even 40 minutes before the cardiac arrhythmia occurred. This would provide sufficient lead time to take appropriate measures, especially in hospitals.

The researchers also hope to further refine their early warning system. Perhaps one day it will be possible to indicate even earlier and more reliably that heart problems may occur in the future. This would mean exercise and diet as well as appropriate medication instead of hospitalization.

Incidentally, the smartwatch in the picture is the Huawei Watch Fit 2 from our review.