Scholarlink

<p>One of the main causes of death globally still is sudden cardiac arrest. Thus, quick defibrillation is desperately needed to raise survival rates. However, the kind of arrhythmia involved determines most of the success of defibrillation. Although shockable rhythms—such as ventricular fibrillation (VF) and pulseless ventricular tachycardia (VT)—demand immediate defibrillation, non-shockable rhythms like asystole and pulmonary electrical activity (PEA) demand different treatment approaches. In this work, we integrate variational mode decomposition (VMD) with deep learning approaches to present a fresh approach for shockable arrhythmias identification. VMD allows the breakdown of electrocardiogram (ECG) signals into several components, each with varying frequency ranges, so enhancing the information accessible for study. Deep learning models are able to differentiate between shockable and non-shockable conditions and then use these degraded signals for training. We investigated several designs, including long short-term memory (LSTM) networks, convolutional neural networks (CNNs), and a hybrid LSTM-CNN. The models were taught using ECG data from the publicly accessible PhysioBank database. We split the ECG recordings into sections of 2, 5, and 8&#xa0;seconds to evaluate performance. Using 5-second segments, the LSTM-CNN hybrid among the tested models obtained the best accuracy—99.50%—on the test set. These encouraging findings imply that combining VMD with deep learning is a potent strategy for precisely spotting shockable arrhythmias. Faster and more focused interventions during cardiac crises made possible by such developments help to improve patient outcomes by themselves. We split training and testing data 70–30. We also compared our technique to other current methods for shockable rhythm detection, and the results amply show the potency and dependability of our approach.</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

Enhancing shockable arrhythmia detection through variational mode decomposition and deep learning: a hybrid LSTM-CNN approach

  • 期刊论文

  • 发表日期：

    2026年2月17日

查看全文
  • Sujata Pedada,
  • Gangula Rajeswara Rao,
  • Jagadeesh B

摘要

One of the main causes of death globally still is sudden cardiac arrest. Thus, quick defibrillation is desperately needed to raise survival rates. However, the kind of arrhythmia involved determines most of the success of defibrillation. Although shockable rhythms—such as ventricular fibrillation (VF) and pulseless ventricular tachycardia (VT)—demand immediate defibrillation, non-shockable rhythms like asystole and pulmonary electrical activity (PEA) demand different treatment approaches. In this work, we integrate variational mode decomposition (VMD) with deep learning approaches to present a fresh approach for shockable arrhythmias identification. VMD allows the breakdown of electrocardiogram (ECG) signals into several components, each with varying frequency ranges, so enhancing the information accessible for study. Deep learning models are able to differentiate between shockable and non-shockable conditions and then use these degraded signals for training. We investigated several designs, including long short-term memory (LSTM) networks, convolutional neural networks (CNNs), and a hybrid LSTM-CNN. The models were taught using ECG data from the publicly accessible PhysioBank database. We split the ECG recordings into sections of 2, 5, and 8 seconds to evaluate performance. Using 5-second segments, the LSTM-CNN hybrid among the tested models obtained the best accuracy—99.50%—on the test set. These encouraging findings imply that combining VMD with deep learning is a potent strategy for precisely spotting shockable arrhythmias. Faster and more focused interventions during cardiac crises made possible by such developments help to improve patient outcomes by themselves. We split training and testing data 70–30. We also compared our technique to other current methods for shockable rhythm detection, and the results amply show the potency and dependability of our approach.