Arrhythmias & electrophysiologyDiagnostic or prognostic study
Sampling frequency has architecture-dependent effects on electrocardiogram deep learning classification: Convolutional neural networks are robust, whereas transformers benefit modestly
- Analysis of the PTB-XL dataset evaluating convolutional neural network (CNN) and Patch Transformer architectures at varying ECG sampling frequencies
- multilabel classification of 5 diagnostic superclasses
- 100, 300, and 500 Hz
Authors' conclusion
ECG sampling frequency had modest, architecture-dependent effects on deep learning classification performance. CNNs were robust across 100-500 Hz, whereas the Patch Transformer benefited incrementally from higher temporal resolution at greater computational cost. These findings support a task- and architecture-aware approach to ECG acquisition in resource-constrained settings.
Study context
- Population
- 12-lead electrocardiograms (ECGs) from the PTB-XL dataseteligible 12-lead ECGs
- Participants
- 15,000 participants
- Comparison
- ECG sampling frequency across CNN and Patch Transformer architectures
Performance
As reported in the abstract
Macro receiver operating characteristic area under the curve
| Test or model | 100 Hz | 300 Hz | 500 Hz |
|---|---|---|---|
| CNN | 0.895 | 0.886 | 0.881 |
| Patch Transformer | 0.856 | 0.864 | 0.877 |
Further findings
Computational cost (training time, memory use, inference latency)
- Higher sampling frequencies increased computational cost for both architectures, with longer training times, greater memory use, and slower inference
Sensitivity and learning-curve analyses
- Supported architecture-dependent trends across the tested configurations
Original publication
Sampling frequency has architecture-dependent effects on electrocardiogram deep learning classification: Convolutional neural networks are robust, whereas transformers benefit modestly. Heart Rhythm. Published 4 August 2026. doi:10.1016/j.hrthm.2026.07.045
Summarised from the published abstract; every number is checked against it