HealthDay News — An artificial intelligence (AI)-enabled electrocardiograph (ECG) acquired during normal sinus rhythm can identify individuals with atrial fibrillation, according to a study published online Aug. 1 in The Lancet.
Zachi I. Attia, from the Mayo Clinic in Rochester, Minnesota, and colleagues developed an AI-enabled ECG using a convolutional neural network to detect the electrocardiographic signature of atrial fibrillation present during normal sinus rhythm. Data were included for 180,922 patients with 649,931 normal sinus rhythm ECGs for analysis, which were allocated to the training, internal validation, and testing data sets in a 7:1:2 ratio.
The researchers found that in the testing data set, 8.4 percent of patients had verified atrial fibrillation before the normal sinus rhythm ECG tested by the model. Atrial fibrillation was identified by a single AI-enabled ECG with an area under the receiver operating characteristic curve (AUC) of 0.87, with sensitivity and specificity of 79.0 and 79.5 percent, respectively, and with an F1 score of 39.2 percent and overall accuracy of 79.4 percent. The AUC was increased to 0.90 with inclusion of all ECGs acquired during the first month of each patient’s window of interest, while sensitivity and specificity increased to 82.3 and 83.4 percent, respectively, the F1 score increased to 45.4 percent, and overall accuracy increased to 83.3 percent.
“This result could have important implications for atrial fibrillation screening and for the management of patients with unexplained stroke,” the authors write.