Scholarlink

<p>Heart disease remains the most prevalent chronic disorder and serves as the primary cause of high mortality worldwide. In recent years, heart disease prediction relying on an Electrocardiogram (ECG) using Deep learning (DL) has gained more attention due to the remarkable success. However, the existing DL methods fail to capture the subtle features and increase the computational complexity resulting in subpar performance. Hence, a novel Tangentially Incremental learning-enabled eXplainable Pyramid-based Scalar invariant Deep Network(TIXPSDN) is proposed in this research for achieving efficient heart disease prediction. Specifically, the Optimized Soft Wavelet Transform (OSWT) is adopted to pre-process the input ECG signal enhancing the quality by mitigating the outliers. Further, the Tangentially Luminant Spiral Movement Optimization (TLuSpMO) adaptively fine-tunes the hyperparameters of the proposed model and eliminates the local convergence issues. In addition, the usage of the incremental distributed learning technique along with the scalar invariant and Kernelized Channel Attention (KCA) mechanism stipulated the learning ability by achieving better scalability and interpretability characteristics. Moreover, the extraction of the pertinent features through the PFVS descriptor enhanced the overall prediction performance. Experimental results report that the proposed method established high efficiency by achieving a high accuracy of 97.81%, F1-score of 96.81, NPV of 96.86%, PPV of 98.46%, error rate of 0.021, and recall of 98.13% with 70% of training using MIT-BIH Arrhythmia database outperforming the other competing algorithms.</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

Advancing Heart Disease Prediction from ECG Signal Using Tangentially Incremental Learning-Based Pyramid Network

  • 期刊论文

  • 发表日期：

    2025年8月4日

查看全文
  • Smita Samrat Mande,
  • Dhanashri Wategaonkar

摘要

Heart disease remains the most prevalent chronic disorder and serves as the primary cause of high mortality worldwide. In recent years, heart disease prediction relying on an Electrocardiogram (ECG) using Deep learning (DL) has gained more attention due to the remarkable success. However, the existing DL methods fail to capture the subtle features and increase the computational complexity resulting in subpar performance. Hence, a novel Tangentially Incremental learning-enabled eXplainable Pyramid-based Scalar invariant Deep Network(TIXPSDN) is proposed in this research for achieving efficient heart disease prediction. Specifically, the Optimized Soft Wavelet Transform (OSWT) is adopted to pre-process the input ECG signal enhancing the quality by mitigating the outliers. Further, the Tangentially Luminant Spiral Movement Optimization (TLuSpMO) adaptively fine-tunes the hyperparameters of the proposed model and eliminates the local convergence issues. In addition, the usage of the incremental distributed learning technique along with the scalar invariant and Kernelized Channel Attention (KCA) mechanism stipulated the learning ability by achieving better scalability and interpretability characteristics. Moreover, the extraction of the pertinent features through the PFVS descriptor enhanced the overall prediction performance. Experimental results report that the proposed method established high efficiency by achieving a high accuracy of 97.81%, F1-score of 96.81, NPV of 96.86%, PPV of 98.46%, error rate of 0.021, and recall of 98.13% with 70% of training using MIT-BIH Arrhythmia database outperforming the other competing algorithms.