New System Predicts Atrial Fibrillation with 83% Accuracy Minutes Before Onset
A novel development, based on artificial intelligence, can now forecast the onset of atrial fibrillation a common form of heart rhythm irregularity half an hour ahead of time.
According to a study, it's sufficient to train the AI on pulse rate data for it to accurately predict the occurrence of atrial fibrillation. The technology, intended to be integrated into smartphones, will analyze data from smartwatches and, if necessary, issue warnings based on its analysis.
The study, published in the journal Patterns, was conducted by researchers at the University of Luxembourg. It documents the training process of the AI, highlighting its potential in rapid intervention to revert the irregular heartbeat back to normal sinus rhythm. This may involve medication or electrical cardioversion.
Predicting such an episode in advance opens the door to early interventions, potentially increasing the chances of recovery for patients, as reported by New Atlas.
The AI model named WARN was trained using data collected from 350 patients at a Chinese hospital, thereby gaining insight into different conditions both normal sinus rhythm and the phenomena preceding and accompanying atrial fibrillation. Drawing from these insights, the model assesses the likelihood of danger.
To identify the signs, researchers used the changes between the R-wave intervals on an EKG (RRI) as a data source.
The model processed 30-second RRI samples at 15-second intervals to calculate the probability of atrial fibrillation. Alongside the main dataset, two validation sets were also used, involving 33 patients each.
The WARN model managed to predict atrial fibrillation with 83% and 73% accuracy, on average 31 and 33 minutes before its onset, respectively.
The goal is for the solution to process data from a smartwatch through a smartphone application, informing the user about their heart rhythm and whether there might be a looming issue.
The practical deployment of this technology is yet to be announced.