Scholarlink

Cardiovascular diseases (CVD) represent a significant global health concern necessitating swift and precise diagnosis, with arrhythmia being a notable example. It results from abnormal heart rhythms and ranks as a primary cause of global morbidity and mortality. To this end the electrocardiogram (ECG/EKG) is a pivotal tool for cardiac diagnosis but requires specialized expertise. In recent research, machine learning (ML) algorithms have advanced heart disease diagnostics by analyzing complex ECG signals. Motivated by this our study aimed to categorize five types of arrhythmias following AAMI EC57 standards, utilizing ML models like random forest, logistic regression, and k-nearest neighbor, as well as deep learning models (DL) including deep neural networks, convolutional neural networks, and stacked autoencoders. Additionally, we applied a majority voting ensemble approach using ML and DL models. Remarkably, DL-based ensemble models consistently outperformed alternative methods, offering a more effective arrhythmia classification approach.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Cardiac Disease Diagnosis with Machine Learning (ML) and Deep Learning (DL): A Comparative Study with Electrocardiogram (ECG/EKG) Data

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Pinaki Ranjan Das,
  • Amiya Kumar Samanta,
  • Tushar Kanti Bera

摘要

Cardiovascular diseases (CVD) represent a significant global health concern necessitating swift and precise diagnosis, with arrhythmia being a notable example. It results from abnormal heart rhythms and ranks as a primary cause of global morbidity and mortality. To this end the electrocardiogram (ECG/EKG) is a pivotal tool for cardiac diagnosis but requires specialized expertise. In recent research, machine learning (ML) algorithms have advanced heart disease diagnostics by analyzing complex ECG signals. Motivated by this our study aimed to categorize five types of arrhythmias following AAMI EC57 standards, utilizing ML models like random forest, logistic regression, and k-nearest neighbor, as well as deep learning models (DL) including deep neural networks, convolutional neural networks, and stacked autoencoders. Additionally, we applied a majority voting ensemble approach using ML and DL models. Remarkably, DL-based ensemble models consistently outperformed alternative methods, offering a more effective arrhythmia classification approach.