A clinical study on Atrial Fibrillation, Premature Ventricular Contraction, and Premature Atrial Contraction screening based on an ECG deep learning model

• DOI: 10.1016/j.asoc.2022.109213
• Publication Date: 2022-06-29T01:26:37Z
• Authors: Jianyuan Hong; Hua-Jung Li; Chung-chi Yang; Chih-Lu Han; Jui-chien Hsieh
• Keywords: Premature atrial contraction

It is still a challenge to develop an electrocardiography (ECG) interpreter based on ECG basic characteristics because of the uncertainty delineation. Based clinical investigation in this study, devices generated interpretations Atrial Fibrillation (AF), Premature Ventricular Contraction (PVC), and (PAC) have high ratios false-positive errors. An interpretation gap exists between cardiologists. This study aimed improve performance AF, PVC, PAC ECGs. first adopted deep learning model delineate features such as P, QRS, T waves 1160 8–10-s lead I or II signals from clinically-used 12-lead device whose AF training dataset. Second, sliding window with 3-RR intervals length applied raw examine delineated window, then determined experiences The results indicate following: (1) delineator achieves good P-, QRS-, T- wave delineation sensitivity/specificity 0.94/0.98, 1.00/0.99, 0.97/0.98, respectively, 48 10-s test ECGs mixed true-positive (2) As compared ECG-device interpretations, precision detection was increased 0.77 0.86, 0.76 0.84, 0.82 0.87 188 Finally, (3) F1 measure, which measure accuracy data but takes false-negative into account, were 0.92, 0.91, 0.83, respectively. In conclusion, overcomes difficulties P-wave discrimination are not documented well previous research improves devices' interpretation. We believe that can facilitate applications ECG, bridge machines'