https://doi.org/10.1140/epjs/s11734-025-01781-y
Regular Article
ECG-ResViT: a hybrid CNN-ViT model for efficient ECG signal classification
School of Electronics Engineering, Vellore Institute of Technology, 600127, Chennai, India
a
sathiyanarayanan.s@vit.ac.in
Received:
24
April
2025
Accepted:
27
June
2025
Published online:
10
July
2025
Electrocardiogram (ECG) analysis is a critical tool in the early detection of life-threatening cardiac conditions, such as Arrhythmia (ARR), Congestive Heart Failure (CHF), and Normal Sinus Rhythm (NSR). While deep learning (DL) models, especially Convolutional Neural Networks (CNNs), have demonstrated success in identifying local patterns in ECG signals, their limited capacity to capture long-range dependencies often impacts performance. On the other hand, Vision Transformers (ViTs), known for their strength in modeling global relationships, are underexplored in the context of 1D biomedical signals. This study introduces ECG-ResViT, a novel hybrid DL architecture that combines the local feature extraction power of CNNs—enhanced with residual blocks and dilated convolutions—with the global attention modeling of ViTs. The architecture leverages positional encoding and multi-head self-attention to effectively capture temporal dependencies in ECG signals. Evaluated on benchmark datasets comprising ARR, CHF, and NSR classes, the proposed model achieves an accuracy of 99.13% and ROC-AUC of 99.94%, outperforming several existing state-of-the-art methods. By balancing accuracy, interpretability, and computational efficiency, ECG-ResViT demonstrates strong potential for real-time clinical deployment and highlights the value of hybrid approaches in advancing biomedical signal classification.
Copyright comment Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
© The Author(s), under exclusive licence to EDP Sciences, Springer-Verlag GmbH Germany, part of Springer Nature 2025
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
