Lead-Level ECG Quality Annotations from Multiple Wearable Devices

61,172 ten-second ECG recordings from 4,540 participants provide a resource for evaluating wearable electrocardiogram data quality. Collected by Jiawei Luo using Holter monitors, ECG vests, and portable ECG cards, the dataset includes fine-grained, lead-level annotations classified by cardiovascular specialists. It was published on figshare in April 2026.

Use Cases

• Train models to classify lead-level ECG signal quality into four grades using annotated 10-second recordings.
• Develop multi-device generalization models using data from Holter monitors, ECG vests, and portable ECG cards.
• Research signal quality under real-world conditions spanning rest and daily activities for remote monitoring applications.

Strengths

• 61,172 ECG recordings provide a substantial sample for model training.
• Fine-grained, lead-level quality annotations by specialists offer precise labels.
• Data from three device types supports research on multi-device generalization.

Limitations

• Geographic origin of the 4,540 participants is unspecified, which may limit generalizability.
• The dataset focuses on signal quality assessment, not specific cardiac diagnoses or events.

Provenance

Source

Jiawei Luo via figshare.

Collection Method

ECG recordings collected using dynamic Holter monitors, wearable ECG vests, and portable ECG cards.

Freshness

Last updated in April 2026.