Scholarlink

<p>An electrocardiogram (ECG) monitors the heart’s electrical activity to assess its function. Cardiac arrhythmias, the irregular heart rhythms, can lead to severe conditions like stroke. ECG recordings are essential for diagnosing these arrhythmias. Due to the critical importance of cardiac arrhythmias, there has been a strong focus on automated classification and identification. Conventional methods need improvement in accurate classification and computational efficiency. This research proposes a bi-directional gated recurrent unit (Bi-GRU) network classifier for arrhythmia detection from ECG signals, utilizing features extracted and encoded by an autoencoder (AE). The dual-tree complex wavelet transform (DTCWT) is used to preprocess the input data, effectively removing noise and baseline. The model is assessed and generalised utilizing two publically accessible MIT-BIH Arrhythmia datasets. The study attains an overall accuracy of 99.06% and 99.4%, surpassing existing state-of-the-art methods. Furthermore, the effectiveness of the proposed model is assessed using several metrics, including precision, recall, and F1-score. Furthermore, post hoc explainability was incorporated through SHAP to elucidate the decision-making process of the model, thus increasing clinical trust by identifying the most significant portions of ECG waveforms in the prediction of arrhythmias.</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

Wavelet Bi-GRUNet: An Explainable Efficient Hybrid Model for Cardiac Arrhythmia Classification

  • 期刊论文

  • 发表日期：

    2026年1月7日

查看全文
  • Priya Mishra,
  • Subham Kumar Padhy

摘要

An electrocardiogram (ECG) monitors the heart’s electrical activity to assess its function. Cardiac arrhythmias, the irregular heart rhythms, can lead to severe conditions like stroke. ECG recordings are essential for diagnosing these arrhythmias. Due to the critical importance of cardiac arrhythmias, there has been a strong focus on automated classification and identification. Conventional methods need improvement in accurate classification and computational efficiency. This research proposes a bi-directional gated recurrent unit (Bi-GRU) network classifier for arrhythmia detection from ECG signals, utilizing features extracted and encoded by an autoencoder (AE). The dual-tree complex wavelet transform (DTCWT) is used to preprocess the input data, effectively removing noise and baseline. The model is assessed and generalised utilizing two publically accessible MIT-BIH Arrhythmia datasets. The study attains an overall accuracy of 99.06% and 99.4%, surpassing existing state-of-the-art methods. Furthermore, the effectiveness of the proposed model is assessed using several metrics, including precision, recall, and F1-score. Furthermore, post hoc explainability was incorporated through SHAP to elucidate the decision-making process of the model, thus increasing clinical trust by identifying the most significant portions of ECG waveforms in the prediction of arrhythmias.