Abstract Body (Do not enter title and authors here): INTRODUCTION: The analysis of Holter ECG recordings on a large scale typically demands substantial clinical resources. This investigation aimed to evaluate whether an artificial intelligence (AI)-powered Holter analysis platform could interpret ambulatory ECG recordings with comparable accuracy to that of a board-certified cardiologist.

METHODS: A total of 328 30-second ECG rhythm strips were analysed from continuous 14-day ECG patch recordings. The recordings underwent independent analysis by the AI platform (PulseAI, Belfast, United Kingdom), six individual cardiologists, and a consensus panel of three cardiologists which served as the gold standard. The presence or absence of five predetermined cardiac arrhythmias were evaluated: atrial fibrillation/flutter (AFIB), ventricular bigeminy, ventricular trigeminy, ventricular tachycardia (VT), and high-grade atrioventricular block (AVB).

RESULTS: The comparison between the AI platform's interpretation and that of individual cardiologists revealed no statistically significant differences in sensitivity and specificity (McNemar's Test, p>0.05). For AFIB, the sensitivity was 0.76 for the AI platform compared to 0.74 for individual cardiologists, with specificities of 0.97 and 0.94, respectively. Similarly, for AVB, the sensitivity was 0.85 for the AI platform and 0.83 for individual cardiologists, with specificities of 0.99 for both. For VT, the sensitivity was 0.71 for the AI platform and 0.78 for individual cardiologists, with specificities of 0.98 and 0.99, respectively. For ventricular bigeminy and ventricular trigeminy, the AI platform demonstrated sensitivities of 0.82 and 0.85, respectively, compared to 0.81 and 0.79 for individual cardiologists, with specificities of 0.99 for all comparisons.

CONCLUSIONS: These results suggest that use of an AI-based platform for Holter ECG analysis can provide rapid ECG interpretation from ambulatory ECG monitors at levels equivalent to those of a board-certified cardiologist. Implementation of this AI-platform could facilitate precise, scalable, and consistent arrhythmia analysis when monitoring patients for arrhythmias.