Sampling frequency has architecture-dependent effects on electrocardiogram deep learning classification: Convolutional neural networks are robust, whereas transformers benefit modestly

Authors

• Cole Freedman, The University of Texas Rio Grande ValleyFollow
• Syed Zawahir Hassan, University of Texas at Austin
• Maximo Alvarez, The University of Texas Rio Grande ValleyFollow
• Peter Elhajj, The University of Texas Rio Grande ValleyFollow
• Paul S. Kim, The University of Texas Rio Grande ValleyFollow

Document Type

Article

Publication Date

7-17-2026

Abstract

Background: Electrocardiograms (ECGs) are deployed across a wide range of environments and frequently operate at varying sampling frequencies, directly influencing bandwidth consumption, storage demands, and battery life.

Objective: This study aims to evaluate the impact of ECG sampling frequency on both classification performance and computational cost across convolutional neural network (CNN) and Patch Transformer architectures.

Methods: Using the PTB-XL dataset, 15,000 eligible 12-lead ECGs were analyzed at 100, 300, and 500 Hz. A CNN and a Patch Transformer were trained for multilabel classification of 5 diagnostic superclasses: normal, myocardial infarction, ST/T-wave change, conduction disturbance, and hypertrophy. Performance was assessed along with training time, memory use, and inference latency.

Results: CNN performance was relatively stable across sampling frequencies, with macro receiver operating characteristic area under the curve values of 0.895, 0.886, and 0.881 at 100, 300, and 500 Hz, respectively. In contrast, the Patch Transformer showed modest improvement with increasing frequency, with macro receiver operating characteristic area under the curve increasing from 0.856 at 100 Hz to 0.864 at 300 Hz and 0.877 at 500 Hz. 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.

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.

Comments

Copyright © 2026 Heart Rhythm Society. Published by Elsevier Inc. All rights reserved.

Recommended Citation

Freedman, C., Hassan, S. Z., Alvarez, M., Elhajj, P., Kim, P., Park, S. K., Park, C., Hirji, S., Grimm, J. C., Vallabhajosyula, S., Nanna, M., Okoh, A., Unlu, O., Pantlin, P., Fenrich, A., & Hong, M. (2026). Sampling frequency has architecture-dependent effects on electrocardiogram deep learning classification: Convolutional neural networks are robust, whereas transformers benefit modestly. Heart rhythm, S1547-5271(26)02657-3. Advance online publication. https://doi.org/10.1016/j.hrthm.2026.07.045  

Creative Commons License

This work is licensed under a Creative Commons Attribution-NonCommercial-No Derivative Works 4.0 International License.

Publication Title

Heart Rhythm

DOI

10.1016/j.hrthm.2026.07.045

Academic Level

medical student