Enhancing cardiovascular disease classification through attention-based deep learning methods

Qianqian Tu; Zejun Yan; Yan Li

doi:10.1117/12.3113507

Enhancing cardiovascular disease classification through attention-based deep learning methods

Qianqian Tu
, Zejun Yan
, Yan Li^*

^*Corresponding author for this work

School of Biological Science and Medical Engineering

Beihang University

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

The field of electrocardiography (ECG) monitoring and diagnosis systems has experienced notable advancements in software and hardware innovations, particularly in the realm of automatic ECG classification algorithms. In this study, we propose a novel Squeeze-and-Excitation ResNet and Transformer encoder model (RSETd) that integrates gender and age demographic features to classify seven categories, including normal and six abnormalities. We also explore an extended version of RSETd to classify 39 categories. Our model demonstrates satisfactory accuracy and time-cost performance: The proposed model achieves an impressive F1-score of 0.779, with a time-cost of 10.862ms per record. To enhance the classification capabilities further, we expand the detection classes to 39 categories of abnormalities, resulting in an F1-score of 0.732, with a time-cost of 104.561ms per record.

Original language	English
Title of host publication	International Conference on Computational Intelligence and Image Analysis, ICCIIA 2026
Editors	Yan Qiang, Petr Hajek, Han Wang
Publisher	SPIE
ISBN (Electronic)	9798902325512
DOIs	https://doi.org/10.1117/12.3113507
State	Published - 13 May 2026
Event	2026 International Conference on Computational Intelligence and Image Analysis, ICCIIA 2026 - Taiyuan, China Duration: 30 Jan 2026 → 1 Feb 2026

Publication series

Name	Proceedings of SPIE - The International Society for Optical Engineering
Volume	14244
ISSN (Print)	0277-786X
ISSN (Electronic)	1996-756X

Conference

Conference	2026 International Conference on Computational Intelligence and Image Analysis, ICCIIA 2026
Country/Territory	China
City	Taiyuan
Period	30/01/26 → 1/02/26

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being

Keywords

deep learning
ECG classification
Squeeze-and-Excitation
Transformer

Access to Document

10.1117/12.3113507

Cite this

Tu, Q., Yan, Z., & Li, Y. (2026). Enhancing cardiovascular disease classification through attention-based deep learning methods. In Y. Qiang, P. Hajek, & H. Wang (Eds.), International Conference on Computational Intelligence and Image Analysis, ICCIIA 2026 Article 142441I (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 14244). SPIE. https://doi.org/10.1117/12.3113507

@inproceedings{8268432a20f7495192cfccf3460e856f,

title = "Enhancing cardiovascular disease classification through attention-based deep learning methods",

abstract = "The field of electrocardiography (ECG) monitoring and diagnosis systems has experienced notable advancements in software and hardware innovations, particularly in the realm of automatic ECG classification algorithms. In this study, we propose a novel Squeeze-and-Excitation ResNet and Transformer encoder model (RSETd) that integrates gender and age demographic features to classify seven categories, including normal and six abnormalities. We also explore an extended version of RSETd to classify 39 categories. Our model demonstrates satisfactory accuracy and time-cost performance: The proposed model achieves an impressive F1-score of 0.779, with a time-cost of 10.862ms per record. To enhance the classification capabilities further, we expand the detection classes to 39 categories of abnormalities, resulting in an F1-score of 0.732, with a time-cost of 104.561ms per record.",

keywords = "deep learning, ECG classification, Squeeze-and-Excitation, Transformer",

author = "Qianqian Tu and Zejun Yan and Yan Li",

note = "Publisher Copyright: {\textcopyright} 2026 SPIE.; 2026 International Conference on Computational Intelligence and Image Analysis, ICCIIA 2026 ; Conference date: 30-01-2026 Through 01-02-2026",

year = "2026",

month = may,

day = "13",

doi = "10.1117/12.3113507",

language = "英语",

series = "Proceedings of SPIE - The International Society for Optical Engineering",

publisher = "SPIE",

editor = "Yan Qiang and Petr Hajek and Han Wang",

booktitle = "International Conference on Computational Intelligence and Image Analysis, ICCIIA 2026",

address = "美国",

}

Tu, Q, Yan, Z & Li, Y 2026, Enhancing cardiovascular disease classification through attention-based deep learning methods. in Y Qiang, P Hajek & H Wang (eds), International Conference on Computational Intelligence and Image Analysis, ICCIIA 2026., 142441I, Proceedings of SPIE - The International Society for Optical Engineering, vol. 14244, SPIE, 2026 International Conference on Computational Intelligence and Image Analysis, ICCIIA 2026, Taiyuan, China, 30/01/26. https://doi.org/10.1117/12.3113507

Enhancing cardiovascular disease classification through attention-based deep learning methods. / Tu, Qianqian; Yan, Zejun; Li, Yan.
International Conference on Computational Intelligence and Image Analysis, ICCIIA 2026. ed. / Yan Qiang; Petr Hajek; Han Wang. SPIE, 2026. 142441I (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 14244).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Enhancing cardiovascular disease classification through attention-based deep learning methods

AU - Tu, Qianqian

AU - Yan, Zejun

AU - Li, Yan

PY - 2026/5/13

Y1 - 2026/5/13

N2 - The field of electrocardiography (ECG) monitoring and diagnosis systems has experienced notable advancements in software and hardware innovations, particularly in the realm of automatic ECG classification algorithms. In this study, we propose a novel Squeeze-and-Excitation ResNet and Transformer encoder model (RSETd) that integrates gender and age demographic features to classify seven categories, including normal and six abnormalities. We also explore an extended version of RSETd to classify 39 categories. Our model demonstrates satisfactory accuracy and time-cost performance: The proposed model achieves an impressive F1-score of 0.779, with a time-cost of 10.862ms per record. To enhance the classification capabilities further, we expand the detection classes to 39 categories of abnormalities, resulting in an F1-score of 0.732, with a time-cost of 104.561ms per record.

AB - The field of electrocardiography (ECG) monitoring and diagnosis systems has experienced notable advancements in software and hardware innovations, particularly in the realm of automatic ECG classification algorithms. In this study, we propose a novel Squeeze-and-Excitation ResNet and Transformer encoder model (RSETd) that integrates gender and age demographic features to classify seven categories, including normal and six abnormalities. We also explore an extended version of RSETd to classify 39 categories. Our model demonstrates satisfactory accuracy and time-cost performance: The proposed model achieves an impressive F1-score of 0.779, with a time-cost of 10.862ms per record. To enhance the classification capabilities further, we expand the detection classes to 39 categories of abnormalities, resulting in an F1-score of 0.732, with a time-cost of 104.561ms per record.

KW - deep learning

KW - ECG classification

KW - Squeeze-and-Excitation

KW - Transformer

UR - https://www.scopus.com/pages/publications/105041104098

U2 - 10.1117/12.3113507

DO - 10.1117/12.3113507

M3 - 会议稿件

AN - SCOPUS:105041104098

T3 - Proceedings of SPIE - The International Society for Optical Engineering

BT - International Conference on Computational Intelligence and Image Analysis, ICCIIA 2026

A2 - Qiang, Yan

A2 - Hajek, Petr

A2 - Wang, Han

PB - SPIE

T2 - 2026 International Conference on Computational Intelligence and Image Analysis, ICCIIA 2026

Y2 - 30 January 2026 through 1 February 2026

ER -

Enhancing cardiovascular disease classification through attention-based deep learning methods

Abstract

Publication series

Conference

UN SDGs

Keywords

Access to Document

Other files and links

Fingerprint

Cite this