Clinical applicability of an artificial intelligence prediction algorithm for early prediction of non-persistent atrial fibrillation

• DOI: 10.3389/fcvm.2023.1168054
• Publication Date: 2023-09-13T21:51:50Z
• Authors: Yeji Kim; Gihun Joo; Bo-Kyung Jeon; Dong-Hyeok Kim; Tae Young Shin; Hyeonseung Im; Junbeom Park
• Keywords: Single Center

Background and aims It is difficult to document atrial fibrillation (AF) on ECG in patients with non-persistent (non-PeAF). There limited understanding of whether an AI prediction algorithm could predict the occurrence non-PeAF from information normal sinus rhythm (SR) a 12-lead ECG. This study aimed derive precise predictive model for screening using SR within 4 weeks. Methods retrospective cohort included aged 18 99 standard (10 seconds) Ewha Womans University Medical Center 3 years. Data were preprocessed into three window periods (which are defined duration detection) – 1 week, 2 weeks, weeks AF detection prospectively. For experiments, we adopted Residual Neural Network based 1D-CNN proposed previous study. We used 7,595 ECGs (extracted 215,875 ECGs) analysis. Results The showed AUC 0.862 F1-score 0.84 1:4 matched group 1-week period. 2-week period, it 0.864 0.85. Finally, 4-week 0.842 0.83. Conclusion possibility risk stratification early non-PeAF. Moreover, this that short period also sufficient detect