Scholarlink

<p>The most serious cardiac consequence is myocardial infarction (MI), which happens when the circulation of blood to the cardiovascular system is entirely or partially restricted. The most recent tool for determining the severity of MI is electrocardiography (ECG). Manual examination of MI-induced ECG alterations is an arduous task. In this work, we develop a deep learning-based system for the automated screening of MI subjects using ECG data. T-F spectral images are extracted from ECG data using the short-time Fourier transform. To build an accurate MI screening system, obtained images are fed into a suggested lightweight deep learning framework using a data-augmentation approach with a 10-fold validation strategy. Developed system obtained an accuracy of 98.19%, the specificity of 98.57%, and sensitivity of 98.86%. The findings show that the suggested approach successfully gathers deep characteristics from the input spectrum. The created T-F pictures and deep learning-based framework can be installed in the cloud and utilized to automatically detect MI patients, potentially improving medical diagnosis performance and reducing health workers’ workload.</p>

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

Time-Frequency Images and Deep Learning-Based Automated Myocardial Infarction Detection System for Remote Monitoring

  • 期刊论文

  • 发表日期：

    2025年10月17日

查看全文
  • Shailesh Khaparkar,
  • Agya Mishra

摘要

The most serious cardiac consequence is myocardial infarction (MI), which happens when the circulation of blood to the cardiovascular system is entirely or partially restricted. The most recent tool for determining the severity of MI is electrocardiography (ECG). Manual examination of MI-induced ECG alterations is an arduous task. In this work, we develop a deep learning-based system for the automated screening of MI subjects using ECG data. T-F spectral images are extracted from ECG data using the short-time Fourier transform. To build an accurate MI screening system, obtained images are fed into a suggested lightweight deep learning framework using a data-augmentation approach with a 10-fold validation strategy. Developed system obtained an accuracy of 98.19%, the specificity of 98.57%, and sensitivity of 98.86%. The findings show that the suggested approach successfully gathers deep characteristics from the input spectrum. The created T-F pictures and deep learning-based framework can be installed in the cloud and utilized to automatically detect MI patients, potentially improving medical diagnosis performance and reducing health workers’ workload.