Scholarlink

<p>Heart failure is the primary cause of mortality and morbidity affecting millions of people worldwide. Early diagnosis and treatment is essential to minimize the risk of complications and management of heart failure patients. Over the years, numerous methods have been developed for the heart failure detection to improve the diagnosis process. However, the existing techniques have some drawbacks, including the necessity for manual intervention, overfitting problems, and a lack of interpretability. Consequently, this research proposes the Butterfly swarming algorithm fused statistical attention and temporal attention-based reconstruction module-enabled Convolutional Neural Network (BfS-FSTReCN) to overcome these restrictions and improve the heart failure detection accuracy using Electrocardiogram (ECG) signals. The BfS-FSTReCN framework exploits the benefits of the fused statistical attention and temporal attention mechanisms, which improve the detection performance as well as minimize the complexity challenges. Specifically, the statistical attention facilitates the model to focus on the statistical features of data for highlighting the important regions based on their distributional characteristics. Besides, the temporal attention allows the model to focus on the dynamic features based on the temporal structure significantly suppressing the redundant information and boosting the overall detection performance. In addition, the reconstruction module with hybrid activation function improves the learning capabilities of the Bfs-FSReCN model to capture the intricate patterns in the input data resulting in enhanced performance. Further, the Butterfly swarming algorithm is employed to optimize the BfS-FSTReCN model’s parameters improving the overall training process. Moreover, the BfSA-optimized feature selection enables the framework to focus on more informative features, thus enhancing the heart failure detection. The experimental results demonstrate the efficiency of the BfS-FSTReCN model in terms of accuracy of 96.83%, specificity of 95.53%, and sensitivity of 98.37%, respectively, for 80% of the training.</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

Butterfly Swarming-Optimized Attention Guided Deep Learning Model for ECG-Based Heart Failure Detection

  • 期刊论文

  • 发表日期：

    2025年11月5日

查看全文
  • Amit Maheshchandra Sahu,
  • Jayant P. Mehare

摘要

Heart failure is the primary cause of mortality and morbidity affecting millions of people worldwide. Early diagnosis and treatment is essential to minimize the risk of complications and management of heart failure patients. Over the years, numerous methods have been developed for the heart failure detection to improve the diagnosis process. However, the existing techniques have some drawbacks, including the necessity for manual intervention, overfitting problems, and a lack of interpretability. Consequently, this research proposes the Butterfly swarming algorithm fused statistical attention and temporal attention-based reconstruction module-enabled Convolutional Neural Network (BfS-FSTReCN) to overcome these restrictions and improve the heart failure detection accuracy using Electrocardiogram (ECG) signals. The BfS-FSTReCN framework exploits the benefits of the fused statistical attention and temporal attention mechanisms, which improve the detection performance as well as minimize the complexity challenges. Specifically, the statistical attention facilitates the model to focus on the statistical features of data for highlighting the important regions based on their distributional characteristics. Besides, the temporal attention allows the model to focus on the dynamic features based on the temporal structure significantly suppressing the redundant information and boosting the overall detection performance. In addition, the reconstruction module with hybrid activation function improves the learning capabilities of the Bfs-FSReCN model to capture the intricate patterns in the input data resulting in enhanced performance. Further, the Butterfly swarming algorithm is employed to optimize the BfS-FSTReCN model’s parameters improving the overall training process. Moreover, the BfSA-optimized feature selection enables the framework to focus on more informative features, thus enhancing the heart failure detection. The experimental results demonstrate the efficiency of the BfS-FSTReCN model in terms of accuracy of 96.83%, specificity of 95.53%, and sensitivity of 98.37%, respectively, for 80% of the training.