II导联心电图中心肌梗死检测与定位：基于多尺度残差模块融合改进通道注意力模型

doi:10.12122/j.issn.1673-4254.2025.08.22

摘要/Abstract

摘要：

目的提高心肌梗死（MI）检测和定位准确性，为临床诊断提供辅助决策支持。方法本文提出了一种基于多尺度残差模块融合改进通道注意力模型（MSF-RB-MCA）。该模型利用II导联心电图（ECG）信号检测和定位MI，通过多尺度残差模块提取不同层次的特征信息，并引入改进通道注意力自动调整特征权重，增强模型对MI区域的关注能力，从而提高MI检测与定位的精度。结果使用公开的PTB数据集对提出的模型进行了5折交叉验证。在MI检测任务中，模型在测试集上的准确率、特异性、敏感度分别达到99.96%、99.84%和99.99%；在MI定位任务中，准确率、特异性、敏感度分别为99.81%、99.98%和99.65%。检测和定位结果均优于其他几种模型。结论本研究提出的MSF-RB-MCA模型在基于II导联ECG信号的MI检测与定位方面表现出色，展现出其在可穿戴设备领域中的广泛应用前景。

关键词: 心肌梗死, 深度学习, 多尺度, 残差模块, 改进通道注意力

Abstract:

Objective We propose a myocardial infarction (MI) detection and localization model for improving the diagnostic accuracy for MI to provide assistance to clinical decision-making. Methods The proposed model was constructed based on multi-scale field residual blocks fusion modified channel attention (MSF-RB-MCA). The model utilizes lead II electrocardiogram (ECG) signals to detect and localize MI, and extracts different levels of feature information through the multi-scale field residual block. A modified channel attention for automatic adjustment of the feature weights was introduced to enhance the model's ability to focus on the MI region, thereby improving the accuracy of MI detection and localization. Results A 5-fold cross-validation test of the model was performed using the publicly available Physikalisch-Technische Bundesanstalt (PTB) dataset. For MI detection, the model achieved an accuracy of 99.96% on the test set with a specificity of 99.84% and a sensitivity of 99.99%. For MI localization, the accuracy, specificity and sensitivity were 99.81%, 99.98% and 99.65%, respectively. The performances of the model for MI detection and localization were superior to those of other comparison models. Conclusion The proposed MSF-RB-MCA model shows excellent performance in AI detection and localization based on lead II ECG signals, demonstrating its great potential for application in wearable devices.

Key words: myocardial infarction, deep learning, multi-scale, residual block, modified channel attention

吴秋岑, 卢学麒, 温耀棋, 洪永, 吴煜良, 陈超敏. II导联心电图中心肌梗死检测与定位：基于多尺度残差模块融合改进通道注意力模型[J]. 南方医科大学学报, 2025, 45(8): 1777-1790.

Qiucen WU, Xueqi LU, Yaoqi WEN, Yong HONG, Yuliang WU, Chaomin CHEN. A myocardial infarction detection and localization model based on multi-scale field residual blocks fusion with modified channel attention[J]. Journal of Southern Medical University, 2025, 45(8): 1777-1790.

图/表 28

参考文献 37

[1]	World Health Organization. Cardiovascular diseases (CVDs)[EB/OL]. [2021-12-26]. . doi：10.1002/9781118910573.ch5
[2]	Acharya UR, Fujita H, Oh SL, et al. Application of deep convolutional neural network for automated detection of myocardial infarction using ECG signals[J]. Inf Sci, 2017, 415: 190-8. doi：10.1016/j.ins.2017.06.027
[3]	Cho Y, Kwon JM, Kim KH, et al. Artificial intelligence algorithm for detecting myocardial infarction using six-lead electrocardiography[J]. Sci Rep, 2020, 10(1): 20495. doi：10.1038/s41598-020-77599-6
[4]	Xiong P, Lee SM, Chan G. Deep learning for detecting and locating myocardial infarction by electrocardiogram: a literature review[J]. Front Cardiovasc Med, 2022, 9: 860032. doi：10.3389/fcvm.2022.860032
[5]	Jian JZ, Ger TR, Lai HH, et al. Detection of myocardial infarction using ECG and multi-scale feature concatenate[J]. Sensors: Basel, 2021, 21(5): 1906. doi：10.3390/s21051906
[6]	Liu XW, Wang H, Li ZJ, et al. Deep learning in ECG diagnosis: a review[J]. Knowl Based Syst, 2021, 227: 107187. doi：10.1016/j.knosys.2021.107187
[7]	Cai JX, Sun WW, Guan JF, et al. Multi-ECGNet for ECG arrythmia multi-label classification[J]. IEEE Access, 2020, 8: 110848-58. doi：10.1109/access.2020.3001284
[8]	Bhaskar NA. Performance analysis of support vector machine and neural networks in detection of myocardial infarction[J]. Procedia Comput Sci, 2015, 46: 20-30. doi：10.1016/j.procs.2015.01.043
[9]	Al-Naima FM, Ali AH, Mahdi SS. Data acquisition for myocardial infarction classification based on wavelets and Neural Networks[C]//2008 5th International Multi-Conference on Systems, Signals and Devices. July 20-22, 2008. Amman. IEEE, 2008: 1-6. doi：10.1109/ssd.2008.4632817
[10]	Sun L, Lu Y, Yang K, et al. ECG analysis using multiple instance learning for myocardial infarction detection[J]. IEEE Trans Biomed Eng, 2012, 59(12): 3348-56. doi：10.1109/tbme.2012.2213597
[11]	Liu B, Liu JK, Wang GQ, et al. A novel electrocardiogram parameterization algorithm and its application in myocardial infarction detection[J]. Comput Biol Med, 2015, 61: 178-84. doi：10.1016/j.compbiomed.2014.08.010
[12]	Chang PC, Lin JJ, Hsieh JC, et al. Myocardial infarction classification with multi-lead ECG using hidden Markov models and Gaussian mixture models[J]. Appl Soft Comput, 2012, 12(10): 3165-75. doi：10.1016/j.asoc.2012.06.004
[13]	Cao D, Lin D, Lv Y. ECG codebook model for Myocardial Infarction detection[C]//2014 10th International Conference on Natural Computation (ICNC). IEEE, 2014: 797-801. doi：10.1109/icnc.2014.6975939
[14]	Lu HL, Ong K, Chia P. An automated ECG classification system based on a neuro-fuzzy system[C]//Computers in Cardiology 2000. Vol.27 (Cat. 00CH37163). Cambridge, MA, USA. IEEE, 2000: 387-390.
[15]	Zhang DD, Yang S, Yuan XH, et al. Interpretable deep learning for automatic diagnosis of 12-lead electrocardiogram[J]. iScience, 2021, 24(4): 102373. doi：10.1016/j.isci.2021.102373
[16]	Jun TJ, Nguyen HM, Kang D, et al. ECG arrhythmia classification using a 2-D convolutional neural network[EB/OL]. 2018: 1804.06812. . doi：10.48550/arXiv.1804.06812
[17]	Kiranyaz S, Ince T, Gabbouj M. Real-time patient-specific ECG classification by 1-D convolutional neural networks[J]. IEEE Trans Biomed Eng, 2016, 63(3): 664-75. doi：10.1109/tbme.2015.2468589
[18]	Al Rahhal MM, Bazi Y, Almubarak H, et al. Dense convolutional networks with focal loss and image generation for electrocardiogram classification[J]. IEEE Access, 2019, 7: 182225-37. doi：10.1109/access.2019.2960116
[19]	Zhang JS, Lin F, Xiong P, et al. Automated detection and localization of myocardial infarction with Staked sparse autoencoder and TreeBagger[J]. IEEE Access, 2019, 7: 70634-42. doi：10.1109/access.2019.2919068
[20]	Ahmad Z, Tabassum A, Guan L, et al. ECG heartbeat classification using multimodal fusion[J]. IEEE Access, 2021, 9: 100615-26. doi：10.1109/access.2021.3097614
[21]	Goldberger AL, Amaral LA, Glass L, et al. PhysioBank, PhysioToolkit, and PhysioNet: components of a new research resource for complex physiologic signals[J]. Circulation, 2000, 101(23): E215-20. doi：10.1161/01.cir.101.23.e215
[22]	Singh BN, Tiwari AK. Optimal selection of wavelet basis function applied to ECG signal denoising[J]. Digit Signal Process, 2006, 16(3): 275-87. doi：10.1016/j.dsp.2005.12.003
[23]	Pan JP, Tompkins WJ. A real-time QRS detection algorithm[J]. IEEE Trans Biomed Eng, 1985, BME-32(3): 230-6. doi：10.1109/tbme.1985.325532
[24]	Miotto R, Wang F, Wang S, et al. Deep learning for healthcare: review, opportunities and challenges[J]. Brief Bioinform, 2018, 19(6): 1236-46. doi：10.1093/bib/bbx044
[25]	Fei NY, Gao YZ, Lu ZW, et al. Z-score normalization, hubness, and few-shot learning[C]//2021 IEEE/CVF International Conference on Computer Vision (ICCV). October 10-17, 2021. Montreal, QC, Canada. IEEE, 2021: 142-151. doi：10.1109/iccv48922.2021.00021
[26]	Zhang X, Lin MJ, Hong Y, et al. MSFT: a multi-scale feature-based transformer model for arrhythmia classification[J]. Biomed Signal Process Control, 2025, 100: 106968. doi：10.1016/j.bspc.2024.106968
[27]	Le KH, Pham HH, Nguyen TBT, et al. LightX3ECG: a lightweight and eXplainable deep learning system for 3-lead electrocardiogram classification[J]. Biomed Signal Process Control, 2023, 85: 104963. doi：10.1016/j.bspc.2023.104963
[28]	Cheng JY, Zou QX, Zhao YX. ECG signal classification based on deep CNN and BiLSTM[J]. BMC Med Inform Decis Mak, 2021, 21(1): 365. doi：10.1186/s12911-021-01736-y
[29]	Sandler M, Howard A, Zhu ML, et al. MobileNetV2: inverted residuals and linear bottlenecks[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. June 18-23, 2018. Salt Lake City, UT. IEEE, 2018: 4510-4520. doi：10.1109/cvpr.2018.00474
[30]	Kattenborn T, Leitloff J, Schiefer F, et al. Review on Convolutional Neural Networks (CNN) in vegetation remote sensing[J]. ISPRS J Photogramm Remote Sens, 2021, 173: 24-49. doi：10.1016/j.isprsjprs.2020.12.010
[31]	Targ S, Almeida D, Lyman K. Resnet in resnet: generalizing residual architectures[EB/OL]. 2016: 1603.08029. .
[32]	He KM, Zhang XY, Ren SQ, et al. Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). June 27-30, 2016. Las Vegas, NV, USA. IEEE, 2016: 770-778. doi：10.1109/cvpr.2016.90
[33]	Hou QB, Zhou DQ, Feng JS. Coordinate attention for efficient mobile network design[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). June 20-25, 2021. Nashville, TN, USA. IEEE, 2021: 13708-13717. doi：10.1109/cvpr46437.2021.01350
[34]	Hu J, Shen L, Sun G. Squeeze-and-excitation networks[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. June 18-23, 2018. Salt Lake City, UT. IEEE, 2018: 7132-7141. doi：10.1109/cvpr.2018.00745
[35]	Woo S, Park J, Lee JY, et al. CBAM: convolutional block attention module[C]//Computer Vision – ECCV 2018. Cham: Springer, 2018: 3-19. doi：10.1007/978-3-030-01234-2_1
[36]	Van der Maaten L, Hinton G. Visualizing data using t-SNE[J]. J Machine Learning Res, 2008, 9(11)：1305-12.
[37]	Feng K, Pi XT, Liu HY, et al. Myocardial infarction classification based on convolutional neural network and recurrent neural network[J]. Appl Sci, 2019, 9(9): 1879. doi：10.3390/app9091879

Class	Number of subjects
AMI	17
ALMI	16
ASMI	27
IMI	30
ILMI	23
IPMI	1
IPLMI	8
LMI	1
PMI	1
PLMI	2
Unknown	22
Total	148

Class	Number of subjects
AMI	17
ALMI	16
ASMI	27
IMI	30
ILMI	23
IPMI	1
IPLMI	8
LMI	1
PMI	1
PLMI	2
Unknown	22
Total	148

Type of disease	Number of patients	Number of heartbeats	Percentage (%)
MI	148	49179	82.53
HC	52	10409	17.47
Total	200	59588	100

Type of disease	Number of patients	Number of heartbeats	Percentage (%)
MI	148	49179	82.53
HC	52	10409	17.47
Total	200	59588	100

Branch1	Branch2	Branch3
Input (ECG signals)
Conv1d (kernel=7, Number of kernel=64)
BN
ReLU
MaxPool1d (kernel=3, Number of kernel=2)
Backbone Block
RB+MCA (Number of kernel=64)	RB+MCA (Number of kernel=64)	RB+MCA (Number of kernel=64)
RB+MCA (Number of kernel=128)	RB+MCA (Number of kernel=128)	RB+MCA (Number of kernel=128)
RB+MCA ( Number of kernel=256)	RB+MCA (Number of kernel=256)	RB+MCA (Number of kernel=256)
RB+MCA ( Number of kernel=512)	RB+MCA (Number of kernel=512)	RB+MCA (Number of kernel=512)
GAP	GAP	GAP
Fusion
Concatenation
Flatten
Dropout (P=0.2)
FC