Data and software credits

Used for

Training and testing. The test ECGs are the tenth fold. The photos made by computer and the printed test pages are made from its recordings.

Credit

• Wagner, P., Strodthoff, N., Bousseljot, R., Samek, W., & Schaeffter, T. PTB-XL, a large publicly available electrocardiography dataset (version 1.0.3). PhysioNet.
• Wagner, P., Strodthoff, N., Bousseljot, R.-D., Kreiseler, D., Lunze, F. I., Samek, W., & Schaeffter, T. PTB-XL: A Large Publicly Available ECG Dataset. Scientific Data.

Printed on every page made from this dataset: Signal: PTB-XL 1.0.3, CC BY 4.0, anonymised. No personal data.

What TraceRead changed

Resampled from 500 Hz to 100 Hz for some tests, masked to printout layouts, drawn as paper pages by computer, printed or turned into photos made by computer, then read again.

Used for

The yardstick for measurements: the published heart rate, PR, QRS, QT and QTc values from two reference ECG programs. Only the published values are used. The programs themselves are not used.

Credit

• Strodthoff, N., Mehari, T., Nagel, C., Aston, P., Sundar, A., Graff, C., Kanters, J., Haverkamp, W., Doessel, O., Loewe, A., Bär, M., & Schaeffter, T. PTB-XL+, a comprehensive electrocardiographic feature dataset (version 1.0.1). PhysioNet.
• Strodthoff, N., Mehari, T., Nagel, C., et al. PTB-XL+, a comprehensive electrocardiographic feature dataset. Scientific Data 10, 279.

What TraceRead changed

None. The published values are used as they are. The PTB-XL credit above applies to this dataset too.

Used for

Testing only, never training: 45,152 ECGs from China that TraceRead AI never saw, including the 176 labelled ST elevation.

Credit

• Zheng, J., Guo, H., & Chu, H. A large scale 12 lead electrocardiogram database for arrhythmia study (version 1.0.0). PhysioNet.
• Zheng, J., Chu, H., Struppa, D., Zhang, J., Yacoub, S. M., El-Askary, H., Chang, A., Ehwerhemuepha, L., Abudayyeh, I., Barrett, A. S., Fu, G., Yao, H., Li, D., Guo, H., & Rakovski, C. Optimal Multi-Stage Arrhythmia Classification Approach. Scientific Reports, 10.

What TraceRead changed

Resampled to 100 Hz for some tests. Labels mapped to PTB-XL statements by TraceRead. 1,002 merged header lines repaired by rule, with the signals untouched. Records rewritten as plain 500 Hz files and drawn as paper pages by computer for the ST elevation test.

Used for

Extra training examples for six labels, and a test of how a drifting baseline hurts a model.

Credit

• CODE-15%: a large scale annotated dataset of 12 lead ECGs (version 1.0.0). Zenodo. Author list as on the Zenodo record.
• Ribeiro, A. H., et al. The article that describes CODE-15%. Nature Communications 11, 1760.
• Lima, E. M., et al. Deep neural network estimated electrocardiographic age as a mortality predictor. medRxiv.

What TraceRead changed

Resampled from 400 Hz to 500 Hz (and to 100 Hz for some tests). Padding removed.

Used for

The truth set for measurements: 200 ECGs with the start and end of every wave marked by cardiologists in every lead.

Credit

• Kalyakulina, A., Yusipov, I., Moskalenko, V., Nikolskiy, A., Kosonogov, K., Zolotykh, N., & Ivanchenko, M. Lobachevsky University Electrocardiography Database (version 1.0.1). PhysioNet.
• Kalyakulina, A. I., Yusipov, I. I., Moskalenko, V. A., Nikolskiy, A. V., Kosonogov, K. A., Osipov, G. V., Zolotykh, N. Yu., & Ivanchenko, M. V. LUDB: A New Open-Access Validation Tool for Electrocardiogram Delineation Algorithms. IEEE Access 8, 186181 to 186190.

What TraceRead changed

None. Used unchanged.