[1]张金龙,杨湘,陈艳红.基于改进丢弃和双通道融合的心肌梗死分类模型[J].计算机技术与发展,2025,(05):121-128.[doi:10.20165/j.cnki.ISSN1673-629X.2025.0001]
　ZHANG Jin-long,YANG Xiang,CHEN Yan-hong.A Myocardial Infarction Classification Model Based on Improved Dropout and Dual Channel Fusion[J].,2025,(05):121-128.[doi:10.20165/j.cnki.ISSN1673-629X.2025.0001]
点击复制

基于改进丢弃和双通道融合的心肌梗死分类模型()

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2025年05期
页码:
121-128
栏目:
人工智能
出版日期:
2025-05-10

文章信息/Info

Title:
A Myocardial Infarction Classification Model Based on Improved Dropout and Dual Channel Fusion
文章编号:
1673-629X(2025)05-0121-08
作者:
张金龙1; 杨湘1; 陈艳红2
1. 武汉科技大学 计算机科学与技术学院,湖北 武汉 430065;
2. 香港大学深圳医院 心律失常科,广东 深圳 518053
Author(s):
ZHANG Jin-long1; YANG Xiang1; CHEN Yan-hong2
1. School of Computer Science and Technology,Wuhan University of Science and Technology,Wuhan 430065,China;
2. Department of Arrhythmias,University of Hong Kong – Shenzhen Hospital,Shenzhen 518053,China
关键词:
心肌梗死分类; 12导联心电图; 双通道融合; 自适应丢弃
Keywords:
myocardial infarction classification; 12 lead electrocardiogram/12 lead ECG; dual channel fusion; adaptive discard
分类号:
TP181;R542.2
DOI:
10.20165/j.cnki.ISSN1673-629X.2025.0001
摘要:
心肌梗死(Myocardial Infarction,MI)是心血管疾病中危及生命的急症,早期检测和准确分类至关重要。 12 导联心电图从多个角度监测心脏的电活动,对心肌梗死的分类具有重要影响。 然而,现有分类模型在处理 12 导联信息冗余所带来的问题时,提出的解决方案考虑不足,会导致重要导联和特征信息的丢失。 此外,这些方法未充分考虑12 导联与心肌梗死之间的关联性,或仅关注单导联的相关性,会导致模型与 MI 的关联不够紧密。 针对以上问题,该文提出了一种基于改进丢弃和双通道融合的心肌梗死分类模型。 该模型采用改进后的自适应丢弃算法(DroSet),有效解决导联所带来的冗余信息问题。 同时,结合医学知识和基础理论,分析导联和心肌梗死的敏感性和关联性,并对其划分成主导心肌梗死导联和辅助心肌梗死导联。 实验结果表明,该模型在 PTBXL 数据集上的分类准确率达到了99.15 % ,证明了模型不仅具有出色的分类性能,还适用于实际应用中的心肌梗死定位判断。
Abstract:
Myocardial Infarction ( MI ) is a life - threatening emergency in cardiovascular disease. Early detection and accurate classification are crucial. 12 lead ECG monitors the electrical activity of the heart from multiple perspectives,which has an important influence on the myocardial infarction classification. However,when the existing classification models deal with the problems caused by 12 lead information redundancy,the proposed solutions are not considered enough,which will lead to the loss of important leads and char-acteristic information. In addition,these methods do not fully consider the correlation between 12 lead and myocardial infarction,or only focus on the correlation of single lead,which will lead to the model and MI are not closely related. To solve the above problems,we propose a myocardial infarction classification model based on improved discarding and dual channel fusion. The model uses an improved adaptive discard algorithm (DroSet) to effectively solve the problem of redundant information caused by leads. At the same time,combined with medical knowledge and basic theory, we analyze the sensitivity and correlation between lead and myocardial infarction,and divide it into leading myocardial infarction lead and auxiliary myocardial infarction lead. The experimental results show that the classification accuracy of the proposed model on ptbxl data set reaches 99. 15% ,which proves that it not only has excellent classification performance,but also is suitable for the location of myocardial infarction in practical application.
更新日期/Last Update: 2025-05-10