Physiology-guided beat-level arrhythmia classification from ECG using a CNN-transformer hybrid neural network

1

TransECG-Net is a hybrid CNN-Transformer network designed for accurate ECG-based arrhythmia classification.
2

The model achieved 99.52% accuracy in classifying five heartbeat classes, outperforming previous models.
3

TransECG-Net utilizes a CNN for local feature extraction and a Transformer for modeling global temporal dependencies.
4

The system maintains low latency of 35 ms and a memory footprint of 28 MB, making it suitable for edge deployment.
5

The framework addresses challenges of noise robustness and privacy-preserving training in federated learning settings.

Frontiers In Cardiovascular Medicine

by Guangfeng Li, Zhidong Zhang, Gang Qiao, Xiaosan Chen, Gangqiang Zhou, Kavimbi Chipusu
July 8, 2026

1