A multicenter study to demonstrate compatibility of ECG-AI acress 12-Lead ECG devices

Electrocardiograms (ECGs) are among the most widely performed diagnostic tests in medicine due to their affordability, accessibility, and utility in detecting cardiovascular disease. However, the diversity of ECG hardware across manufacturers introduces the potential for device-specific variability in signal characteristics, which could affect the performance of downstream ECG-based artificial intelligence (ECG-AI) algorithms. Despite the high sensitivity and specificity of some ECG-AI tools, their clinical applicability may be limited by lack of generalizability across devices. To study the extent of this limitation, we aim to compare the output of multiple ECG-AI models across a range of ECG machines.