Efficiency of artificial intelligence-enabled handheld single-lead device for identifying recurrent atrial fibrillation after radiofrequency ablation
CHEN Laite, JIANG Chenyang
1. School of Medicine, Zhejiang University, Hangzhou Zhejiang 310058; 2. Department of Cardiology, Sir Run Run Shaw Hospital, Affiliated with the Zhejiang University School of Medicine, Hangzhou Zhejiang 310000, China
Abstract：Objective To explore the feasibility of handheld devices for screening recurrent atrial fibrillation, and to assess the efficiency of the configurated artificial intelligence algorithm (AIA) for identifying the recurrence of atrial fibrillation. Methods We selected 67 patients diagnosed with paroxysmal atrial fibrillation and treated by radiofrequency ablation. During one-year follow-up after the operation, the ECG information of patients was collected by handheld single-lead device, and uploaded to the cloud at the onset of dizziness, chest tightness, palpitations, shortness of breath and other symptoms. AIA was utilized to analyze and process the collected ECG signals, and identify the event of atrial fibrillation recurrence. Taking the diagnosis of two ECG doctors as the golden criterion, we calculated the sensitivity, specificity, accuracy rate, positive prediction value and F1 score to assess the efficiency of AIA. Results In the 67 patients with paroxysmal atrial fibrillation, AIA screens out 30 cases with recurrent atrial fibrillation while 25 cases are diagnosed with atrial fibrillation recurrence by ECG doctors. The sensitivity, specificity, accuracy rate, positive prediction value and F1 score of AIA in identifying recurrent atrial fibrillation are separately 83.33%, 100.00%, 88.10%, 92.54% and 0.91. There is a strong consistency between AIA and manual check by ECG doctors in diagnosing recurrent atrial fibrillation (Kappa coefficient=0.85). Conclusion AIA-based handheld singlelead device could be used to identify recurrent atrial fibrillation during the follow up after ablation.
