·One hundred and fourteen subjects were included and underwent ECG examination simultaneously. Subjects were placed in a supine position and upright positions before and after exercise. Wearable dynamic ECG and twelve-lead ECG were used for ECG collection. Using the twelve-lead ECG as standard, the effectiveness of AF detection results, including consistency, sensitivity, specificity, positive predictive value and negative predictive value for detection of AF, was evaluated.

·According to the ECG diagnosis results, the patients were divided into two groups, including 61 patients in the non-AF group and 53 patients in the AF group. The age of the AF group was significantly higher than that of the non-AF group (P=0.000); CHA2DS2-VASc score of the AF group was higher than that of the non-AF group (P=0.001). The proportion of patients with coronary heart disease in the AF group was higher than that in the non-AF group (P=0.014). There were significant differences in the proportion of oral anticoagulants, antiplatelet agents, calcium channel blockers, diuretics, digoxin, and β-blockers between the two groups (all P<0.05). The 60 s-ECG monitoring of the wearable dynamic ECG recorder was automatically determined by the artificial intelligence (AI) algorithm in a supine position: 47 cases of AF, 65 cases of non-AF, and 2 cases not determined. Compared with that of twelve-lead ECG ("unable to judge" as false positive and false negative), the diagnostic consistency of the wearable dynamic ECG recorder for AF was 94.74% (95%CI 88.76%－97.80%). The sensitivity of the wearable dynamic ECG recorder to diagnose AF was 88.68% (95%CI 77.06%－95.07%), and the specific was 100% (95%CI 92.91%－100%). The positive predictive value was 100% (95%CI 90.98%－100%), and the negative predictive value was 91.04% (95%CI 81.48%－96.16%). In an upright position, it was automatically determined by the AI algorithm: 50 AF cases, 61 non-AF cases, and 1 case not determined. Compared with that of the twelve-lead ECG ("unable to judge" as false positive and false negative), the diagnostic consistency of the wearable dynamic ECG recorder for AF was 97.37% (95%CI 92.21%－99.44%). The sensitivity of the wearable dynamic ECG recorder to diagnose AF was 94.34% (95%CI 84.03%－98.65%), and the specific was 100% (95%CI 92.91%－100%). The positive predictive value was 100% (95%CI 91.48%－100%), and the negative predictive value was 95.31% (95%CI 86.57%－98.92%). The results after exercise were the same as that of the standing position. No adverse events occurred throughout the test, and no apparent device defect was found.

·The wearable dynamic ECG recorder has higher specificity and positive predictive value. Moreover, the operation is simple, which could increase the detection rate of paroxysmal AF.