ZZU-ecgQA: A dataset for evaluating the quality of wearable ECG data

In the context of the rapid development of wearable devices and telemedicine, the quality of electrocardiogram (ECG) signals is a critical factor influencing the reliability of subsequent analyses and the accuracy of clinical decision-making. However, existing publicly available ECG signal quality assessment datasets are often limited to single-device acquisitions, controlled environments, or coarse-grained, record-level annotations, which restrict their applicability to real-world, multi-device, and multi-scenario research. To address these limitations, we constructed ZZU-ecgQA, a lead-level ECG signal quality assessment dataset covering multiple devices and scenarios. The dataset comprises 61,172 ten-second ECG recordings from 4,540 participants, collected using three types of devices: dynamic Holter monitors, wearable ECG vests, and portable ECG cards. These recordings span a range of real-world conditions, from rest to daily activities, and include both single-lead and 12-lead configurations. Unlike existing datasets that primarily rely on record-level annotations, ZZU-ecgQA provides fine-grained, lead-level quality annotations, with each lead in every recording classified into four quality grades by cardiovascular specialists based on signal clarity and diagnostic value. This approach preserves usable lead information while minimizing the impact of contaminated leads on overall analyses. The ZZU-ecgQA dataset can support research in ECG signal quality assessment, multi-device generalization modeling, and remote monitoring.